Post on 08-Nov-2021
A randomized air filter intervention study of air pollution
and fetal growth in a highly polluted community:
the Ulaanbaatar Gestation and Air Pollution Research
(UGAAR) study
by
Prabjit Barn
M.Sc., University of British Columbia, 2006
B.Sc., University of British Columbia, 2003
Thesis Submitted in Partial Fulfillment of the
Requirements for the Degree of
Doctor of Philosophy
in the
Doctor of Philosophy Program
Faculty of Health Sciences
© Prabjit Barn 2018
SIMON FRASER UNIVERSITY
Fall 2018
Copyright in this work rests with the author. Please ensure that any reproduction
or re-use is done in accordance with the relevant national copyright legislation.
ii
Approval
Name: Prabjit Barn
Degree: Doctor of Philosophy
Title: A randomized air filter intervention study of air
pollution and fetal growth in a highly polluted
community: the Ulaanbaatar Gestation and Air
Pollution Research (UGAAR) study
Examining Committee: Chair: Tim Takaro
Professor
Ryan W. Allen
Senior Supervisor
Associate Professor
Bruce P. Lanphear
Supervisor
Professor
Patricia A. Janssen
Supervisor
Professor
School of Population and Public Health
Faculty of Medicine
University of British Columbia
Meghan Winters
Internal Examiner
Associate Professor
Amy Padula
External Examiner
Assistant Professor
Reproductive Sciences
University of California San Francisco
Date Defended/Approved: October 1, 2018
iv
Abstract
Background: Gestational exposure to fine particulate matter (PM2.5) and cadmium may
impair fetal growth. Portable high efficiency particulate air (HEPA) filter air cleaners can
reduce indoor PM2.5, but their effect on fetal growth has not been evaluated.
Objectives: We assessed (1) HEPA cleaner effectiveness in reducing residential indoor
PM2.5 and maternal blood cadmium, (2) the effect of HEPA cleaners on fetal growth, and
(3) the relationship between maternal cadmium exposure and fetal growth, among non-
smoking pregnant women in Ulaanbaatar, Mongolia.
Methods: We randomized 540 participants at ≤ 18 weeks gestation to an intervention (1-2
HEPA cleaners in homes from early pregnancy until childbirth) or control (no HEPA
cleaners) group. We collected exposure, health, and demographic data through home and
clinic visits and from clinic records. We measured one-week indoor PM2.5 concentrations
in early (~11 weeks gestation) and late (~31 weeks gestation) pregnancy, collected blood
samples in late pregnancy for analysis of cadmium, and obtained birth data at delivery. We
evaluated the effect of the intervention on our primary outcome, birth weight, and other
fetal growth indicators using unadjusted linear and logistic regression and time-to-event
analysis, in intention-to-treat analyses. We also used multiple linear and logistic regression
to assess the relationships between log2-transformed blood cadmium and fetal growth.
Results: HEPA cleaners reduced indoor PM2.5 and blood cadmium concentrations by 29%
(95% CI: 21, 37%) and 14% (95% CI: 4, 23%), respectively. Among 463 live births, the
median (25th, 75th percentile) birth weights for control and intervention participants were
3450 g (3150, 3800 g) and 3550 g (3200, 3800 g), respectively, but the intervention was
not associated with an increase in birth weight (18 g; 95% CI: -84, 120 g). In a pre-specified
subgroup analysis of 429 term births the intervention was associated with an 85 g (95% CI:
3, 167 g) increase in mean birth weight. A doubling of blood cadmium was associated with
an 86 g (95% CI: 26, 145 g) reduction in birth weight.
Conclusions: Our findings provide further evidence that PM2.5 and cadmium exposures
during pregnancy impair fetal growth and that exposure reduction during pregnancy can
reduce these effects. Portable HEPA cleaners are an effective household-level intervention
but reductions in air pollution emissions are needed to realize the largest public health
benefits.
vii
Acknowledgements
This dissertation is a culmination of work by a large hard-working team. First and foremost,
thank you to Dr. Ryan Allen for dreaming up this ambitious study and for being such a
wonderful supervisor. Throughout the last five years you have been engaged, supportive,
helpful, and so enthusiastic. You also introduced me to the best Indian food in Ulaanbaatar,
for which I will remain eternally grateful. Thank you to Dr. Enkhee Gombojav for
overseeing the day to day work involved with UGAAR, including the large amount of data
collection. To my committee members, Dr. Bruce Lanphear and Dr. Patricia Janssen, thank
you for giving so willingly of your time, expertise, and input. Thank you also to all the co-
authors of our manuscripts, and in particular to Dr. Jennifer Hutcheon - your feedback
made this work better.
My trips to Ulaanbaatar and getting to know the UGAAR study team were the highlights
of my PhD experience. Thank you to the entire team for their hard work and dedication
toward this study, and in particular to Buyan, Bolor, and Dr. Gerel. Also, thank you to the
UGAAR participants who gave so freely of their time and without whom this study would
not be possible. This study was funded by the Canadian Institutes of Health Research
(CIHR) and I was fortunate to receive a CIHR doctoral award.
I am so grateful for my little community. Thank you Emily for always listening and giving
the best advice, Soutomi for your positivity and much needed goofiness, Aman, my fellow
grad student for understanding this journey, Meghan for your random “you are awesome”
texts, Bolor for making Ulaanbaatar feel like home, and Anisha, my thesis angel, for getting
me through the final stretch. My family has been my biggest support. Ma, thank you for
always believing that I am smarter than I really am. Pan, thank you for your unconditional
love, support, and silliness. You inspire me in so many ways and I am so lucky that you
are my sister. Randy, thank you for your constant thoughtfulness. Maja and Phil, thank you
for your support and encouragement. To my husband Aio, thank you for so many things.
viii
Thank you for talking to anyone who would listen about this work, for celebrating each
tiny success, for the many pep talks, and for always believing in me. The last five years
have been full of so many special moments and I could not imagine going through this
journey with anyone else. And finally, to my son Arjun, I love you so much – thank you
for being the best motivation for completing this thesis.
ix
Table of Contents
Approval ............................................................................................................................. ii
Ethics Statement................................................................................................................. iii
Abstract .............................................................................................................................. iv
Dedication .......................................................................................................................... vi
Acknowledgements ........................................................................................................... vii
Table of Contents ............................................................................................................... ix
List of Tables .................................................................................................................... xii
List of Figures .................................................................................................................. xiii
List of Acronyms ............................................................................................................. xiv
Preface............................................................................................................................... xv
Chapter 1. Introduction ................................................................................................. 1
1.1. Background ............................................................................................................... 2
Impaired fetal growth and health ...................................................................... 2
PM2.5 and fetal growth ...................................................................................... 3
Cadmium and fetal growth ............................................................................... 6
Biological mechanisms ..................................................................................... 7
Portable air cleaners .......................................................................................... 7
1.2. Rationale ................................................................................................................. 11
Chapter 2. The effect of portable HEPA filter air cleaners on indoor PM2.5
concentrations and second hand tobacco smoke exposure among pregnant
women in Ulaanbaatar, Mongolia ....................................................................... 13
2.1. Abstract ................................................................................................................... 13
2.2. Introduction ............................................................................................................. 14
2.3. Methods .................................................................................................................. 16
Study population ............................................................................................. 16
Study design ................................................................................................... 17
Data collection ................................................................................................ 18
Indoor residential air pollution measurements .......................................................... 18
Outdoor air pollution data ......................................................................................... 20
Relative humidity ...................................................................................................... 20
Blood cadmium ......................................................................................................... 20
Hair nicotine .............................................................................................................. 20
Other data .................................................................................................................. 21
Data analysis ................................................................................................... 21
2.4. Results ..................................................................................................................... 23
2.5. Discussion ............................................................................................................... 32
2.6. Conclusions ............................................................................................................. 36
x
Chapter 3. The effect of portable HEPA filter air cleaner use during pregnancy on
fetal growth: the UGAAR randomized controlled trial .................................... 38
3.1. Abstract ................................................................................................................... 38
3.2. Introduction ............................................................................................................. 39
3.3. Methods .................................................................................................................. 40
Study design ................................................................................................... 40
Participants ..................................................................................................... 41
Randomization and blinding ........................................................................... 42
Intervention ..................................................................................................... 42
Procedures ...................................................................................................... 42
Outcomes ........................................................................................................ 43
Statistical analysis........................................................................................... 44
Role of funding source ................................................................................... 46
3.4. Results ..................................................................................................................... 46
3.5. Discussion ............................................................................................................... 57
3.6. Conclusions ............................................................................................................. 61
Chapter 4. Gestational cadmium exposure and fetal growth in Ulaanbaatar,
Mongolia ................................................................................................................ 62
4.1. Abstract ................................................................................................................... 62
4.2. Introduction ............................................................................................................. 63
4.3. Methods .................................................................................................................. 65
Data collection ................................................................................................ 65
Blood cadmium concentrations ................................................................................. 65
Fetal growth outcomes .............................................................................................. 66
Determinants of cadmium exposure and co-variates ................................................. 66
Data analysis ................................................................................................... 67
4.4. Results ..................................................................................................................... 68
4.5. Discussion ............................................................................................................... 73
4.6. Conclusions ............................................................................................................. 78
Chapter 5. Discussion................................................................................................... 79
5.1. Summary ................................................................................................................. 79
Effect of portable HEPA cleaners on indoor PM2.5 concentrations and SHS
exposure (Chapter 2) ..................................................................................................... 79
Effect of HEPA cleaner use on fetal growth (Chapter 3) ............................... 80
Effect of gestational cadmium exposure on fetal growth (Chapter 4) ............ 80
5.2. Synthesis and significance of findings.................................................................... 81
5.3. Strengths and limitations ........................................................................................ 85
5.4. Future research directions ....................................................................................... 86
5.5. Conclusions ............................................................................................................. 89
xi
References ........................................................................................................................ 90
Appendix A. Summary of studies investigating maternal cadmium exposure and
fetal growth .......................................................................................................... 110
Appendix B. Supplemental material for chapter two ............................................ 122
Appendix C. Supplemental material for chapter three ......................................... 133
Appendix D. Supplemental material for chapter four ........................................... 137
xii
List of Tables
Table 1.1. Summary of studies investigating reductions in residential indoor
particulate matter associated with portable HEPA cleaner use .................. 9
Table 2.1. Summary of household, personal, and behavioural characteristics by
intervention status ..................................................................................... 26
Table 2.2. Effect of portable HEPA filter air cleaner use on one-week residential
indoor PM2.5 concentrations. ..................................................................... 29
Table 3.1. Summary of baseline characteristics for control and intervention
participants. ............................................................................................... 49
Table 3.2. Summary of variables assessed during pregnancy .................................... 51
Table 3.3. Effect of the intervention on fetal growth and birth outcomes ................. 54
Table 4.1 Maternal blood cadmium concentrations (µg/L) by maternal and newborn
characteristics ............................................................................................ 70
Table 4.2 Effect of a doubling of maternal blood cadmium concentrations on fetal
growth outcomes ....................................................................................... 72
xiii
List of Figures
Figure 1.1. Summary of meta-analyses investigating relationships between PM2.5 and
(a) birth weight, (b) low birth weight, (c) preterm birth, and (d) small for
gestational age, where N is the number of studies included in each meta-
analysis ........................................................................................................ 5
Figure 2.1. Summary of data collection ...................................................................... 18
Figure 2.2. Trial profile ............................................................................................... 24
Figure 2.3. Distribution of one-week indoor PM2.5 concentrations in control and
intervention homes stratified by season and measurement (the first
measurement was made when air cleaners were newly deployed and the
second measurement was made after approximately five months of use). 31
Figure 3.1 Data collection .......................................................................................... 43
Figure 3.2 Trial profile ............................................................................................... 47
Figure 3.3 Distribution of birth weight by treatment assignment for all births (a) and
term births (b). .......................................................................................... 53
Figure 3.4. Effect of the intervention on birth weight in stratified analyses for (a) all
births and (b) term births........................................................................... 56
Figure 4.1 Estimated effects of a doubling of maternal blood cadmium (Cd)
concentrations on birth weight in stratified analyses ................................ 73
xiv
List of Acronyms
GC-MS/MS
GM
HEPA
ICP-MS
IUGR
LBW
LMIC
LOQ
PM
PM2.5
PTB
RCT
RH
SGA
SHS
TEOM
UGAAR
Gas chromatography-tandem mass spectrometry
Geometric mean
High efficiency particulate air
Inductively coupled plasma-mass spectrometry
Intrauterine growth restriction
Low birth weight
Low- and middle-income countries
Limit of quantification
Particulate matter
Fine particulate matter
Preterm birth
Randomized controlled trial
Relative humidity
Small for gestational age
Second hand smoke
Tapered element oscillating microbalance
Ulaanbaatar Gestation and Air Pollution Research
xv
Preface
This thesis is organized into five chapters. Chapter one is an introductory chapter that
provides a background, rationale, and the research questions addressed in this work, and
chapter five is a discussion chapter that provides a synthesis of the research conducted for
this thesis. Chapters two, three, and four are research chapters that were written as
manuscripts for publication. At the time of this thesis submission, chapter two had been
published in Science of the Total Environment and chapter three had been accepted for
publication in Environment International. Both papers are a result of feedback and input
from many co-authors, all of whom have approved the final version of the published
manuscripts.
Prabjit Barn, Enkhjargal Gombojav, Chimedsuren Ochir, Bayarkhuu Laagan, Bolor
Beejin, Gerel Naidan, Buyantushig Boldbaatar, Jargalsaikhan Galsuren,
Tsogtbaatar Byambaa, Craig Janes, Patricia A. Janssen, Bruce P. Lanphear, Tim K.
Takaro, Scott A. Venners, Glenys M. Webster, Weiran Yuchi, Christopher D.
Palmer, Patrick J. Parsons, Young Man Roh, Ryan W. Allen. The effect of portable
HEPA filter air cleaners on indoor PM2.5 concentrations and second hand tobacco
smoke exposure among pregnant women in Ulaanbaatar, Mongolia: The UGAAR
randomized controlled trial. Science of The Total Environment. 2018; 615:1379-89.
Prabjit Barn, Enkhjargal Gombojav, Chimedsuren Ochir, Buyantushig Boldbaatar,
Bolor Beejin, Gerel Naidan, Jargalsaikhan Galsuren, Bayarkhuu Legtseg,
Tsogtbaatar Byambaa, Jennifer A. Hutcheon, Craig Janes, Patricia A. Janssen,
Bruce P. Lanphear, Lawrence C. McCandless, Tim K. Takaro, Scott A. Venners,
Glenys M. Webster, Ryan W. Allen. The effect of portable HEPA filter air cleaner
use during pregnancy on fetal growth: the UGAAR randomized controlled trial.
Environment International. 2018. (Accepted)
xvi
Chapter four has been prepared for submission. As with the previous manuscripts, co-
authors provided feedback and input on the version included in this thesis.
Prabjit Barn, Enkhjargal Gombojav, Chimedsuren Ochir, Buyantushig Boldbaatar,
Bolor Beejin, Gerel Naidan, Jargalsaikhan Galsuren, Bayarkhuu Legtseg,
Tsogtbaatar Byambaa, Jennifer A. Hutcheon, Craig Janes, Patricia A. Janssen,
Bruce P. Lanphear, Lawrence C. McCandless, Tim K. Takaro, Scott A. Venners,
Glenys M. Webster, Christopher D. Palmer, Patrick J. Parsons, and Ryan W. Allen.
Gestational cadmium exposure and fetal growth in Ulaanbaatar, Mongolia.
1
Chapter 1.
Introduction
Fine particulate air pollution (PM2.5) is a leading contributor to morbidity and mortality
world-wide due to its well-established impacts on cardiovascular disease, chronic lung
disease, respiratory infections, and lung cancer.1 In 2016, 95% of the world’s population
lived in areas where outdoor PM2.5 concentrations exceeded the World Health
Organization’s annual average guideline of 10 µg/m3.2 Although concentrations in high-
income countries continue to decrease, largely due to stricter emissions regulations,
improvements in technology, and a shift toward cleaner fuels, the global population-
weighted PM2.5 concentration increased by 18% from 2010 (43 µg/m3) to 2016
(51 µg/m3).2 This increase was largely driven by low- and middle-income countries
(LMIC), which bear the largest burden of the public health impacts of air pollution.3
Recent evidence suggests that exposures very early in life, particularly during fetal
development, can have detrimental effects on health. Observational studies have linked
outdoor PM2.5 concentrations during pregnancy to decreased birth weight and increased
risks of low birth weight, small for gestational age, and preterm birth.4-10 Much of the
research consists of observational studies that link fixed-site outdoor PM2.5 concentrations
to administrative data on fetal growth outcomes. These studies, while informative, are often
limited in their ability to accurately assess exposure and account for confounding variables.
Moreover, most studies have been conducted in high-income countries where air pollution
concentrations are relatively low and pollution sources may be different from those in
LMIC. The biological understanding of how air pollution affects fetal growth is also
incomplete. Finally, few intervention studies have been conducted to understand how air
pollution affects fetal growth. Intervention studies offer opportunities to study risk-
reduction strategies and can strengthen causal links between air pollution and fetal growth.
2
Cadmium, a ubiquitous metal, has also been linked to impaired fetal growth, but the
evidence is less consistent.11-23,24 Several studies have reported larger effects on growth
restriction in girls versus boys.11,13,17,20,22,25 Most studies have focused on diet or tobacco
smoke as the dominant sources of exposure. Sources of airborne cadmium other than
tobacco smoke have not been well studied among pregnant women.
The aims of this thesis were to investigate the impact of portable high efficiency particulate
air (HEPA) filter air cleaner (henceforth “HEPA cleaner”) use on indoor residential PM2.5
and maternal blood cadmium concentrations, as well as on fetal growth, and to study the
relationship between cadmium and fetal growth.
1.1. Background
Impaired fetal growth and health
Birth weight and low birth weight are commonly used indicators of fetal growth. Low birth
weight is typically defined as weighing less than 2,500 g regardless of gestational age.
Birth weight is a relatively easy measure to collect with high precision and validity26 and
is also strongly correlated with health in early and later life.27,28 Decreases in birth weight
can occur due to fetal growth restriction and/or shorter gestational duration. To distinguish
between these pathways, researchers have restricted analyses to term births, typically
defined as ≥37 weeks or categorized fetal growth as small for gestational age or preterm
birth. Small for gestational age is defined as being less than a given cut-off value of birth
weight for gestational age and sex for a given population; commonly used cut-offs include
the 3rd, 5th, or 10th percentiles or two standard deviations less than an average weight, for
a given population.29,30 Preterm birth is defined as delivery occurring <37 weeks of
gestation.
3
Fetal growth impairment and preterm birth elevate the risk of death, disease, and disability
in early and later life.31-33 Growth restricted babies are at increased risk of stillbirth,
childhood obesity, and as adults, of type-2 diabetes, hypertension and coronary heart
disease.26,31-35 Preterm birth is associated with increased risks of neonatal mortality, and
short- and long-term pulmonary and neurological morbidity, including wheeze and asthma
in childhood.36-38 The link between unfavorable intrauterine conditions and health in later
life was first described by Barker and colleagues in what became to be known as the
“Barker hypothesis” and was later expanded to the developmental origins of adult health
and disease hypothesis.39,40 This hypothesis centres on the idea that unfavorable
intrauterine conditions force the fetus to make irreversible adaptions that favor immediate
survival but have lasting effects on organ morphology, vasculature, physiology, endocrine
and metabolic functioning.41,42
PM2.5 and fetal growth
There is growing evidence of a link between air pollution and fetal growth. Over 150
studies and 11 meta-analyses investigating this relationship have been conducted, with
most studies focusing on PM2.5 as the main pollutant of interest.4-10,43-46 The bulk of the
research consists of large observational studies that link data from outdoor government
monitoring networks to administrative data on birth outcomes. More recently, researchers
have attempted to develop more refined measures of exposure using satellite data, land use
regression models, or a combination of assessment methods.47-49 Overall, the bulk of the
current research reports small but relatively consistent effects of PM2.5 on birth weight, low
birth weight, preterm birth, and small for gestational age.4-10,43-45
A recent meta-analysis of 32 observational studies reported a 16 g (95% CI: 5, 27 g)
reduction in birth weight and an odds ratio (OR) of 1.09 (95% CI: 1.03, 1.15) for low birth
weight per 10 µg/m3 increase in PM2.5 over full pregnancy.4 The majority of the studies in
the paper limited their analyses to term births, generally defined as ≥ 37 weeks gestation,
as a measure of “normal” fetal growth. Other meta-analyses have reported similar effects
4
of PM2.5 on birth weight and low birth weight, in addition to increased risks of preterm
birth and small for gestational age (Figure 1.1).5-10,43-45 Exposure in the latter half of
pregnancy may play a particularly important role (Figure 1.1).
5
Figure 1.1. Summary of meta-analyses investigating relationships between PM2.5 and (a) birth weight, (b) low birth weight, (c) preterm birth, and (d) small for gestational age, where N is the number of studies
included in each meta-analysis
6
There are several limitations of the existing research. Most studies have been conducted in
high-income countries where air pollution levels are relatively low and vary little over the
course of pregnancy. Outdoor air pollution concentrations may also not provide an accurate
measure of total exposures since exposures inside the home or in other microenvironments,
such as at work or during a commute, are not considered. Much of the previous research
has also not adequately accounted for factors that could bias interpretation of study
findings, such as maternal smoking, second hand smoke (SHS) exposure, and maternal
nutrition. Finally, few studies have looked at the composition of PM2.5 to understand how
different pollutant mixes may impact toxicity.
Cadmium and fetal growth
Cadmium is a ubiquitous metal found in the environment. Diet and tobacco smoke are
typically the main sources of cadmium among the general public.50 Tobacco smoke is the
largest source of cadmium among smokers51 while SHS is an important source among non-
smokers.52-55 Emissions from coal combustion, waste incineration and battery
manufacturing and recycling may be important sources in some communities,51,56-58 but
have not been adequately studied among pregnant women.
The links between cadmium exposure and health effects such as cancer, and kidney and
bone disease are well-established.50,51 Research also links maternal cadmium exposure
during pregnancy, typically measured as concentrations in blood or urine, to impaired fetal
growth although findings are mixed (Table A.1).11-18,20-25,59-73 The inconsistent findings are
likely due, in part, to differences in study design, sample size, and exposure levels. Several
studies have also reported greater effects in girls versus boys.11,13,17,20,22,25 Most of these
studies have focused on tobacco smoke or diet as the main sources of exposure, but few
studies have been conducted among non-smoking pregnant populations in communities
heavily impacted by coal smoke.17,74,75
7
Biological mechanisms
PM2.5 and cadmium may exert their toxicity on fetal growth through overlapping pathways.
Once inhaled, PM2.5 and cadmium can cause local oxidative stress and inflammation, which
initiates a cascade of systemic oxidative stress and inflammatory responses that not only
cause direct damage to cells but also lead to secondary processes such as endothelial
dysfunction and increases in blood coagulation and blood pressure.76-78 During pregnancy,
these pathways are thought to interfere with the development and functioning of the
placenta, ultimately decreasing the transfer of oxygen and nutrients to the fetus.
Both pollutants may also work through additional direct pathways. PM2.5 may directly
damage DNA by reacting with molecules to form chemical bonds with DNA called
adducts.79 Unrepaired DNA adducts can disrupt gene expression and create abnormal
proteins that can impair cellular growth and differentiation and interfere with the
production of growth factors and hormones.80,81 Cadmium, which accumulates in the
placenta, can impair the transfer of essential nutrients59,82 and the release of progesterone.83
PM2.5 and cadmium may both cause epigenetic changes that alter gene expression without
changing DNA sequence, primarily through DNA methylation, histone modifications, and
microRNA activity.84 Epigenetic mechanisms play a critical role in fetal development by
coordinating apoptosis, cell growth, and cell differentiation.85 During pregnancy, cadmium
has been suggested to cause hypomethylation of growth genes in girls resulting in larger
growth deficits.17
Portable air cleaners
Portable HEPA cleaners are a promising intervention to improve our understanding of the
relationship between PM2.5 and fetal growth and to reduce the potential risks. HEPA
cleaners lower PM2.5 concentrations through mechanical ventilation. By definition HEPA
filters reduce 99.7% of particles sized 0.3 µm, with lower efficiency for larger and smaller
8
particles. The effectiveness of HEPA cleaners at reducing indoor particle concentrations
also depends on the volume of air that can be cleaned by the unit, which in turn is related
to the size of the room in which it is placed, as well as the air exchange in the room or
home. Although portable air cleaners are designed to clean air in a single room, they have
been found to reduce whole house PM2.5 concentrations in some studies.86
Portable HEPA cleaner use has been linked to reductions of 25-79% in indoor residential
PM2.5 concentrations (Table 1.1).87-95 All but two of these studies were conducted in high-
income countries where PM2.5 concentrations were relatively low (<12 µg/m3 for most
studies); little is known about the impact of these air cleaners in highly polluted settings.
Additionally, few studies have tested HEPA cleaner use over periods longer than two
weeks. The effects of portable air cleaners on fetal growth have not been studied, but their
use has been linked to improvements in cardiovascular effects that may be also be
important to fetal growth such as improved endothelial function, reduced systemic
inflammation and decreased blood pressure.88,91,95,96
9
Table 1.1. Summary of studies investigating reductions in residential indoor particulate matter associated with portable
HEPA cleaner use
Study Source(s), location, and study design % Reduction
Zhan et al. 201887 No specific source
Beijing, China
One unit was operated in homes for 48 hrs with the filter in place (filtering period) and
48 hrs without the filter (control period); the filtering period was randomly assigned
(n=5).
79
(from a mean of 49.0 to 8.5
μg/m3)
Shao et al. 201788 No specific source
Beijing, China
One unit was operated in homes for two weeks with the filter in place (filtering period)
and two weeks without the filter (control period); the filtering period was randomly
assigned (n=20).
60
(from a mean of 60.0 to
24.0 μg/m3)
Kajbafzadeh et al.
201589
Wood smoke and traffic Vancouver, Canada
Two units were operated in homes for one week with the filters in place (filtering period)
and one week without the filters (control period); the filtering period was randomly
assigned (n=44).
40
(from a mean of 7.1 to
4.3 μg/m3)
Batterman et al.
201290
No specific source (total suspended particles)
Detroit, US
Homes of asthmatic children were randomly assigned to one of three groups for three to
four consecutive seasons: (i) control (n=37); (ii) one unit (n=47); (iii) one unit and an air
conditioner (n=42).
45
(from mean of 21.4 to
11.8 µg/m3)
Allen et al. 201191 Wood smoke
Smithers, Canada
Two units were operated in homes for one week with the filters in place (filtering period)
and one week without the filters (control period); the filtering period was randomly
assigned (n=25).
60
(from mean of 11.2 to
4.6 μg/m3)
10
Butz et al. 201192 Second hand smoke
Baltimore, US
Homes of children exposed to indoor second hand smoke were randomly assigned to one
of three groups over six months: (i) control (n=44); (ii) two units (n=41); (iii) two units
plus home visits from a health coach (n=41).
47
(from mean of 33.9 to
17.9 μg/m3)
Lanphear et al.
201193
Second hand smoke (particles > 0.3 µm)
Cincinnati, US
Homes of children exposed to indoor second hand smoke were randomly assigned to one
of two groups for one-year months: (i) two sham units (n=115); (ii) two units (n=110).
38 at six months
(from 4.0 x 106/ft3 to
2.5 x 106/ft3)
25 at one year
(from 4.0 x 106/ft3 to
3.0 x 106/ft3
Barn et al. 200894 Wildfire and wood smoke
British Columbia, Canada
One unit was operated in homes for 24 hrs with the filter in place (filtering period) and
24 hrs without the filter (control period); the filtering period was randomly assigned
(n=53).
58
(from mean of 6.7 to
4.2 μg/m3)
Brauner et al. 200895 Traffic
Copenhagen, Netherlands
Two units were operated in homes for 48 hrs with the filter in place (filtering period) and
48 hrs without the filter (control period); the filtering period was randomly assigned
(n=21).
62
(from geometric mean of
12.6 to 4.7 μg/m3)
11
1.2. Rationale
There is growing evidence that air pollution exposures during pregnancy can impair fetal
growth. Outdoor PM2.5 exposure during pregnancy has been linked to small but consistent
reductions in fetal growth. Cadmium, an important component of tobacco and coal smoke,
has also been linked to impaired fetal growth. Few studies have been conducted in highly
polluted settings, and even fewer randomized intervention studies have been conducted to
understand how reductions in air pollution may benefit fetal growth. Given that air
pollution is ubiquitous and increasing in much of the world, and the importance of fetal
growth to health in early and later life, it is important to understand how air pollution
exposure during pregnancy affects fetal growth and the potential benefits of exposure
reduction.
This work investigates the impact of portable HEPA cleaners on fetal growth in
Ulaanbaatar, Mongolia, one of the most polluted cities in the world. The population-
weighted annual average PM2.5 concentration in the city is approximately 70 µg/m3.97 The
high concentrations of air pollution are primarily due to emissions from coal combustion
in the city. Household coal use occurs in neighborhoods of traditional Mongolian felt-lined
dwellings (gers) and poorly constructed one or two-story wood and brick homes. Roughly
60% of the city’s residents live in these neighborhoods. Each ger stove burns an average
of five tons of coal per year, and coal stoves in ger neighborhoods are responsible for 45-
70% of total PM2.5 concentrations in the city.98-100 Air pollution emissions linked to
household coal use are expected to increase further as the population in ger neighbourhoods
increases.101 The remainder of Ulaanbaatar’s residents live in apartments, which receive
electricity from three coal-fired power plants. These power plants and an increasing
number of motor vehicles also contribute to air pollution in the city,102 although the
majority of air pollution comes from gers.99,100 Smoking rates are also high in Mongolia;
nearly 40% of men smoke compared with roughly 7% of women.97,103
12
We conducted a randomized controlled trial in which 540 non-smoking pregnant women
were randomized into an intervention group that received one or two HEPA cleaners to use
in their apartments from enrollment to delivery, or a control group that received no HEPA
cleaners. This thesis explores three research questions, which are addressed in chapters 2,
3, and 4, respectively:
1) To what extent do portable HEPA cleaners reduce indoor PM2.5 and blood cadmium
concentrations in a highly polluted community? (Chapter 2)
2) Can portable HEPA cleaner use during pregnancy improve fetal growth and birth
outcomes in a highly polluted community, as measured by birth weight, length,
head circumference, ponderal index, low birth weight, small for gestational age,
and preterm birth? (Chapter 3)
3) What is the relationship between cadmium exposure during pregnancy and fetal
growth? (Chapter 4)
13
Chapter 2.
The effect of portable HEPA filter air cleaners on indoor
PM2.5 concentrations and second hand tobacco smoke
exposure among pregnant women in Ulaanbaatar,
Mongolia
2.1. Abstract
Background: Portable HEPA filter air cleaners can reduce indoor fine particulate matter
(PM2.5), but their use has not been adequately evaluated in high pollution settings. We
assessed air cleaner effectiveness in reducing indoor residential PM2.5 and second hand
smoke (SHS) exposures among non-smoking pregnant women in Ulaanbaatar, Mongolia.
Methods: We randomized 540 participants to an intervention group receiving 1 or 2 HEPA
filter air cleaners or a control group receiving no air cleaners. We followed 259 intervention
and 253 control participants to the end of pregnancy. We measured one-week indoor
residential PM2.5 concentrations in early (~11 weeks gestation) and late (~31 weeks
gestation) pregnancy and collected outdoor PM2.5 data from centrally-located government
monitors. We assessed blood cadmium in late pregnancy. Hair nicotine was quantified in
a subset (n=125) to evaluate blood cadmium as a biomarker of SHS exposure. We
evaluated air cleaner effectiveness using mixed effects and multiple linear regression
models and used stratified models and interaction terms to evaluate potential modifiers of
effectiveness.
Results: The overall geometric mean (GM) one-week outdoor PM2.5 concentration was
47.9 µg/m3 (95% CI: 44.6, 51.6 µg/m3), with highest concentrations in winter
(118.0 µg/m3; 95% CI: 110.4, 126.2 µg/m3). One-week indoor and outdoor PM2.5
concentrations were correlated (r=0.69). Indoor PM2.5 concentrations were 29% (95% CI:
21, 37%) lower in intervention versus control apartments, with GMs of 17.3 µg/m3 (95%
14
CI: 15.8, 18.8 µg/m3) and 24.5 µg/m3 (95% CI: 22.2, 27.0 µg/m3), respectively. Air cleaner
effectiveness was greater when air cleaners were first deployed (40 %; 95% CI: 31, 48%)
than after approximately five months of use (15%; 95% CI: 0, 27%). Blood cadmium
concentrations were 14% (95% CI: 4, 23%) lower among intervention participants, likely
due to reduced SHS exposure.
Conclusions: Portable HEPA filter air cleaners can lower indoor PM2.5 concentrations and
SHS exposures in highly polluted settings.
2.2. Introduction
Outdoor fine particulate matter (PM2.5) air pollution is a leading global public health risk
factor.1,104 The enormous public health impact of PM2.5 is due in part to the large number
of people exposed. In 2013, 87% of the world’s population lived in areas where PM2.5
concentrations exceeded the World Health Organization annual average guideline of
10 μg/m3.105 Despite decreasing concentrations in many high-income countries, the global
population-weighted PM2.5 concentrations increased by over 20% between 1990 and 2013
due largely to increasing concentrations in Asia.105 PM2.5 is a risk factor for numerous
health conditions including ischemic heart disease, stroke, chronic obstructive pulmonary
disease, cancer, and lower respiratory infections.1,104 A growing body of evidence also links
PM2.5 exposure with impaired fetal growth, an important indicator of health in early
childhood and over the life course.31-33
Reducing PM2.5 concentrations results in substantial public health benefits.106-108 From a
public health perspective, interventions that reduce pollution emissions and exposure
among large populations are generally preferable to those that reduce exposure at the
individual or household level. However, because community-wide improvements in air
quality usually occur over decades,109 it is important to identify interventions that can
15
reduce household exposures in the near term until emissions can be reduced to acceptable
levels.
Portable HEPA filter air cleaners are a promising household level intervention to reduce
indoor PM2.5 concentrations. PM2.5 readily infiltrates into buildings,110-112 so a substantial
portion of exposure to PM2.5 of outdoor origin actually occurs indoors, where individuals
spend the majority of their time.113 Many countries with high outdoor air pollution
concentrations also have a high prevalence of smoking, so air cleaners have the potential
advantage of reducing exposure to both outdoor pollution that infiltrates indoors and
indoor-generated pollution from cigarettes and other sources. Air cleaners are widely
available and relatively inexpensive to purchase and operate.114 Previous studies have
linked portable air cleaner use in residences to reductions of 32-68% in concentrations of
particles from various outdoor and indoor sources, including traffic, wildfire and residential
wood smoke, and second hand tobacco smoke (SHS).89-95,115-117 Much of this work has
been conducted in high-income settings where PM2.5 concentrations and smoking rates are
relatively low,118 so little is known about the efficacy of portable air cleaners in highly
polluted settings. Additionally, most studies of air cleaner use have been conducted over
short periods ranging from a few days to weeks, with few evaluations of efficacy over
longer durations.119
The Ulaanbaatar Gestation and Air Pollution Research (UGAAR) study is a randomized
controlled trial designed to assess the effect of portable HEPA filter air cleaner use during
pregnancy on fetal growth and early childhood development (ClinicalTrials.gov Identifier:
NCT01741051). Our study was conducted in Ulaanbaatar, Mongolia’s capital city, which
is home to roughly one-half of the country’s total population of three million.120
Ulaanbaatar is one of the coldest and most polluted cities in the world. The population-
weighted annual average PM2.5 concentration in the city is approximately 70 µg/m3.97
Ulaanbaatar is located in a valley with mountains to the north and south, which together
with cold temperatures, contribute to inversions that exacerbate the poor air quality in
16
winter. Wintertime PM2.5 emissions are dominated by residential heating with coal.97 Coal
combustion is also linked to other pollutants, including cadmium.121 Household coal use
occurs in ger (a traditional felt-lined Mongolian dwelling) neighbourhoods surrounding the
city where roughly 60 % of the city’s population resides.98 In 2013, there were an estimated
164 000 to 185 000 ger households in the city,101 each burning an average of approximately
five tons of coal per year.98 Air pollution emissions linked to household coal use are
expected to increase further as the population in ger neighbourhoods increases.101 The
remainder of Ulaanbaatar’s residents live in apartments, which receive electricity from
three coal-fired power plants. These power plants and an increasing number of motor
vehicles also contribute to air pollution in the city.102 We have previously estimated that
approximately 10% of the mortality in Ulaanbaatar is attributable to outdoor PM2.5.122 The
objective of this analysis was to quantify the impact of HEPA filter air cleaner use during
pregnancy on indoor residential PM2.5 and blood cadmium concentrations.
2.3. Methods
Study population
Our study population consisted of women in Ulaanbaatar who met the following eligibility
criteria: 18 years or older, in the early stages (≤ 18 weeks) of a single-gestation pregnancy,
non-smoker, living in an apartment, planning to give birth in a maternity hospital in
Ulaanbaatar, and not using an air cleaner in the home at enrollment. Initially, recruitment
of participants was done in coordination with the reproductive health clinic at the
Sukhbaatar district Health Centre in Ulaanbaatar. This city district was targeted due to its
large population living in apartments, its proximity to the ger area north of the city centre,
and our relationships with staff at the district hospital. To increase participant recruitment,
we established a second study office in September 2014 at the first branch location of the
Sukhbaatar Health Centre (see Figure B.1). We excluded women living in gers due to
concerns about the reliability of electricity in ger neighbourhoods and the possibility that
higher air exchange rates in gers would make portable HEPA filter air cleaners ineffective.
17
Study design
We randomly assigned 540 participants to the intervention or control group.
Randomization was done using sealed opaque envelopes containing randomly generated
“filter” or “control” allocations and labelled with participant identification numbers that
ran from one to 580. Allocation was done on a 1:1 ratio. Participants in the intervention
group received one or two portable HEPA filter air cleaners (AP-1009CH, Coway, Korea)
depending on the size of their apartment, and air cleaners were used from the first home
visit until childbirth. Apartments with a total area less than 40 m2 received one air cleaner
and those with areas greater or equal to 40 m2 received two air cleaners. The air cleaners
had a clean air delivery rate for tobacco smoke (particles sized 0.09-1.0 µm) of 149 ft3/m,
which is appropriate for use in rooms up to approximately 22 m2. The commercially
available model has an internal PM sensor and “mood light” that changes colour based on
the PM concentration, but this feature was disabled to avoid biasing the behaviour of
UGAAR participants. The air cleaners used in UGAAR were also modified to operate only
on the second-highest fan setting with an internal timer that counted total hours of use.
Timer data were retrieved once each participant completed the study. Unfortunately, the
internal timers proved to have limited value because initiating the timer required the air
cleaner to be turned on while also pressing specific buttons. Participants were given
instructions on the procedure, but if a participant turned on the air cleaner (e.g., after the
unit was turned off, unplugged, or in the event of a power failure) without initiating the
timer then subsequent air cleaner usage was not logged. For smaller apartments, air
cleaners were placed in the main living area of the home, and for larger apartments, the
second unit was placed in the participant's bedroom. Air cleaners were deployed with new
pre-filters, which help to remove large debris, and HEPA filters. Participants were shown
how to clean the pre-filter, but we did not replace pre-filters or HEPA filters during the
study. Participants were encouraged to use the air cleaners continuously throughout the
study period. The control group received no air cleaners.
18
Data collection
Data collection took place from January 2014 to December 2015. We collected data at
home and clinic visits that occurred in early (5-18 weeks gestation) and late (24-37weeks
gestation) pregnancy (Figure 2.1). We collected air pollution measurements over one-week
periods following the two home visits. Whole blood and hair samples were collected during
the second clinic visit. We administered questionnaires at both clinic visits to collect data
on housing and lifestyle (e.g. SHS exposures, time activity patterns) characteristics.
Participants were compensated with a payment of 65,000 Mongolian tugriks
(approximately $45 Canadian) upon completion of data collection, and a pro-rated amount
was provided to participants who withdrew before completion of the study. The study
protocol was approved by the Simon Fraser University Office of Research Ethics
(2013s0016) and the Ministry of Health Medical Ethics Approval Committee (No.7).
Written consent was obtained from participants prior to their enrollment into the study.
Figure 2.1. Summary of data collection
Indoor residential air pollution measurements
We measured particle number concentrations in all apartments during two one-week
sampling campaigns using Dylos laser particle counters (DC1700; Dylos Corporation,
19
Riverside, California, USA). These instruments quantify particle count concentrations in
two particle size ranges: >0.5 µm and >2.5 µm. The commercially available Dylos
monitors log particle counts at one-minute intervals and display counts in real time, but the
units used in UGAAR were modified to log data at five-minute intervals (to allow one-
week of data to be logged). Real-time particle count displays were disabled to avoid biasing
participants’ behavior. We used the difference between the small and large particle size
counts since it has previously been shown to provide the best approximation of PM2.5
concentrations, with reported Dylos-PM2.5 correlations ranging between 0.55 and
0.99.48,123-127 We conducted co-location tests of all Dylos monitors to identify and
discontinue the use of monitors showing poor performance (see Text B.1).
We collected co-located Dylos particle counts and gravimetric PM2.5 measurements in a
subset of 90 apartments, roughly 20 % of our sample. The data were used to establish the
empirical relationship between Dylos particle counts and PM2.5 mass concentrations (in
units of µg/m3) since this relationship depends on the optical properties of the aerosol being
measured. These apartments represent a convenience sample because although they were
randomly chosen to capture a representative sample of intervention and control apartments
across multiple seasons, measurements were only conducted if participants gave
permission for additional sampling. Gravimetric PM2.5 samples were collected onto 37-
mm Teflon filters using Harvard Personal Environmental Monitors (HPEM; Air
Diagnostics and Engineering, Inc., Harrison, ME) connected to mass flow controlled BGI
400 air pumps (BGI, Inc., Waltham, MA) operated at 4 L/min. Filters were weighted in
triplicate before and after sampling, and the average of the three measurements was taken.
The air pollution sampling equipment was placed in the main activity room, typically on a
table or shelf, as far as possible away from the air cleaner, pollution sources, ventilation
systems, and bright light sources.
20
Outdoor air pollution data
Outdoor PM2.5 concentrations were obtained from two centrally-located government-run
monitoring stations. Measurements were made using tapered element oscillating
microbalance (TEOM) monitors.
Relative humidity
Continuous measurements of relative humidity (RH) were made using HOBO loggers
(ux100-011; Onset Computer Corporation; Bourne, MA, USA) in the subset of apartments
selected for gravimetric PM2.5 monitoring. RH was of interest since it can impact the light
scattering properties of particles, thereby influencing the relationship between Dylos
particle counts and PM2.5 mass concentrations. The Dylos has previously been shown to
record artificially high particle counts when RH exceeds approximately 90%.128
Blood cadmium
Whole blood samples were collected from 382 participants by a nurse at the reproductive
health clinic during the second clinic visit. Samples were refrigerated and shipped to the
Wadsworth Center (New York State Department of Health, Albany, New York, USA) for
analysis within six weeks of collection. Samples were analyzed for cadmium using
quadrupole-based inductively coupled plasma-mass spectrometry (ICP-MS), with matrix-
matched calibration.129 The limit of quantification (LOQ), which was based on US
Environmental Protection Agency recommendations, was 0.043 µg/L. Two samples were
below the LOQ; concentrations of LOQ/2 were assigned to these samples.130
Hair nicotine
Hair samples were collected during second clinic visits for analysis of nicotine, an
indicator of SHS. Approximately 30-50 strands of hair (>30 mg) were cut close to the scalp
at the occipital area of the head. Participants were asked if they had chemically treated their
hair in the previous three months since chemical treatment can affect hair nicotine
concentrations.131 After collection, hair samples were placed into a plastic bag and stored
21
at room temperature before being shipped for analysis to the Clinical Pharmacology
Laboratory at the University of California, San Francisco. Hair samples were used
primarily to evaluate blood cadmium as a biomarker of SHS exposure, so 125 hair samples
were selected for analysis to capture potentially low and high SHS exposures among
intervention and control participants, based on whether participants lived with a smoker.
Samples were additionally limited to participants who had a blood cadmium measurement
and those who did not chemically treat their hair. Four-cm samples were analyzed to
represent SHS exposures occurring in the approximately four months prior to data
collection. Hair samples were washed, digested, and then analyzed by gas chromatography-
tandem mass spectrometry (GC-MS/MS). Five samples were below the LOQ of 0.036
ng/mg; concentrations of LOQ/2 were assigned to these samples.130
Other data
Study technicians conducted a home assessment during the first home visit to determine
the area and volume of each room, and total area of the home. Study technicians also
determined the building location using a global positioning system (GPS). If a participant
moved between visits, study technicians also conducted an assessment during the second
home visit. During both clinic visits, staff administered a questionnaire to obtain
information on health, medical history, and lifestyle factors such as alcohol use, smoking,
and exposure to SHS. We quantified air cleaner use in intervention apartments using
information provided on the questionnaire administered at the second clinic visit.
Specifically, participants were asked to estimate the percentage of time that air cleaner
units were used since they were installed in the home. For apartments with two air cleaners,
we averaged the reported use for both units.
Data analysis
We conducted a series of quality control and data cleaning steps on particle count data prior
to analysis, including removing incomplete data, which resulted in the removal of 464
22
(51 %) one-week Dylos measurements. We assessed baseline housing, personal, and
behavioral characteristics among participants with zero and one or two Dylos
measurements. Although participants with no measurements spent less time at home in
early pregnancy (15.7 hours/day, 95% CI: 15.1, 16.2 hours/day) compared with
participants with one or two measurements (16.3 hours/day, 95% CI: 15.9, 16.8 hours/day,
p=0.02), we found no other significant differences between these groups (Table B.1),
indicating that the participants and homes included in our analysis are representative of the
full UGAAR cohort. The effect of RH on particle count data was determined to be
negligible since hourly RH measured in apartments never exceeded 85%. We found strong
agreement between the one-week particle counts and gravimetric PM2.5 concentrations
(R2=0.94, n=22), and used this relationship to convert Dylos particle counts to mass
concentrations (see Text B.2). We averaged outdoor PM2.5 concentrations measured at the
two monitoring sites and calculated one-week averages corresponding to each week of
indoor PM2.5 monitoring in apartments and examined correlations between indoor and
outdoor concentrations. Potential differences by intervention assignment in baseline
housing, personal and behavioral characteristics were assessed using Fisher’s exact tests,
t-tests, and Mann-Whitney tests as appropriate.
Linear and mixed effects regression models were used to assess the impact of the
intervention on indoor PM2.5 and blood cadmium concentrations. All exposure variables
were log-transformed to improve the normality of model residuals, and results are
presented as percent concentration reductions in the intervention group relative to the
control group. Since one-week indoor PM2.5 concentrations were measured twice for
participants, we assessed the effect of the intervention based on all data using mixed effects
models, and for each visit separately using multiple linear regression. For mixed effects
models, we used an unstructured covariance matrix and entered intervention status as a
fixed effect and apartment (participant) as a random intercept to account for repeated
measurements in apartments. Results of indoor PM2.5 models are shown both unadjusted
and adjusted for outdoor PM2.5 concentrations. All analyses comparing intervention and
control groups were based on randomized intervention assignments, and all analyses
23
involving the number of air cleaners were based on the actual number deployed in homes
by study staff. Standard regression diagnostics were conducted on all models.
To evaluate effect modification, we also ran the regression models after stratifying by
variables that we hypothesized might modify air cleaner effectiveness such as number of
air cleaners, air cleaner density (number of air cleaners per 100 m2 of home area), reported
air cleaner use, season, window opening, living with a smoker, living in a home where
smoking occurred indoors, and, for blood cadmium, time spent indoors at home.
Information on time-dependent variables, such as living with a smoker and time spent at
home, was obtained in both early and late pregnancy. For stratifications involving PM2.5,
we used data collected at both time points. For stratified models of blood cadmium, we
used data collected in late pregnancy to reflect more relevant exposure periods. The half-
life of cadmium in blood ranges from roughly 75-128 days.132 Differences in air cleaner
effectiveness between strata were evaluated using interaction terms in the regression
models.
Finally, we evaluated the role of SHS as a source of cadmium exposure. We calculated
correlations between blood cadmium and hair nicotine and compared concentrations of
both biomarkers between smoking and non-smoking households.
2.4. Results
Five hundred and forty women were recruited at a mean gestation of 10.3 weeks (range:
4.0-18.0 weeks). Two hundred and seventy-two participants were randomized to the
control group and 268 were randomized to the intervention group (Figure 2.2). Eight
participants received incorrect treatments. These participants were retained in the dataset
and analyzed according to their assigned treatment groups. Twenty-eight (5%) participants
were lost to follow up, leaving for analysis 512 participants followed to the end of
24
pregnancy. In total, 236 and 211 one-week PM2.5 concentrations measured in early (first
measurement) and later (second measurement) pregnancy, respectively, were analyzed, as
well as 382 whole blood samples and 125 hair samples. Differences in several
characteristics that might influence exposure to PM2.5 and cadmium were examined among
participants who remained in the study and those who were lost to follow up. No significant
differences were found for housing characteristics such as area of home, age of home,
window usage, as well as other characteristics, such as time spent at home, living with a
smoker and season of enrollment into the study (see Table B.2). Participants lost to follow
up were more likely to use a non-UGAAR study air cleaner (i.e. not provided by the study;
p=0.05).
Figure 2.2. Trial profile
25
At baseline, control and intervention participants had similar home ages, total home areas,
and window opening behaviour (Table 2.1). Approximately half of the participants in both
groups reported living with a smoker at any time in pregnancy, and 8% of participants in
both groups reported smoking at any time during pregnancy. Control and intervention
participants spent on average 16 hours per day indoors at home in both early and late
pregnancy. The majority of participants (80%) reported working outside the home during
pregnancy More participants in the control group changed address (9%) compared with the
intervention group (5%). Control participants were also more likely to live in apartments
located on lower floors (56%) than intervention participants (46%). The number of
participants enrolled into the study each season was similar for both groups, with the
highest enrollment occurring during winter and fall. Among the intervention group, 70
households received one air cleaner and 186 households received two air cleaners. Air
cleaner density, calculated as the number of air cleaners per total area of the home, was
similar for apartments with one and two area cleaners, with geometric means of 3.0 air
cleaners/100 m2 (95% CI: 2.8, 3.2 air cleaners/100 m2) and 2.9 air cleaners/100 m2 (9% CI:
2.7, 3.1 air cleaners/100 m2), respectively. Air cleaners were reported to be used for a
geometric mean of 64% of the study duration, and use did not differ by the number of air
cleaners deployed. Thirteen participants in the control group and seven in the intervention
group reported using a non-UGAAR air cleaner.
26
Table 2.1. Summary of household, personal, and behavioural characteristics by
intervention status
Control Group
(n = 253)
Intervention Group
(n = 259) p-value
GM (95 % CI)
or N %
GM (95 % CI)
or N %
Housing characteristics
Total home area (m2) 52.3 (48.7, 55.8) 94 54.6 (51.2, 58.2) 97 0.20
Not recorded 6 3
Age of home (years) 10.6 (8.6, 13.1) 66 11.2 (9.4, 13.3) 73 0.96
Not recorded 34 27
Window opening in winter
Open < half the month 118 47 130 50 0.50
Open ≥ half the month 129 51 126 49
Not recorded 6 2 3 1
Window opening in summer
Open < half the month 24 9 35 13 0.17
Open ≥ half the month 224 89 222 86
Not recorded 5 2 2 1
Outdoor PM2.5 (µg/m3) 54.3 (50.5 58.4) 96 55.0 (51.6, 58.6) 99 0.85
Not available 4 1
Personal and behavioral characteristics
Week of pregnancy at
enrollment into the study 9.9 (9.6, 10.2) 100 9.9 (9.6, 10.3)
10
0
0.94
Season of enrollment into the study
Winter (Dec-Feb) 89 35 78 30 0.46
Spring (Mar-May) 72 28 70 27
Summer (Jun-Aug) 27 11 35 14
Fall (Sep-Nov) 65 26 76 29
Time spent indoors at home in
early pregnancy (hours/day) 16.0 (15.5, 16.5) 79 16.3 (15.9, 16.8) 74
0.41
Not recorded 21 26
Time spent indoors at home in
late pregnancy (hours/day) 15.6 (14.9, 16.3) 47 15.8 (15.1, 16.5) 60
0.69
Not recorded 53 40
Lived with a smoker at any time in pregnancy
No 118 47 131 51 0.47
Yes 127 50 123 47
Not recorded 8 3 5 2
Smoking occurred in the home at any time in pregnancy
No 173 53 176 56 0.70
Yes 73 29 81 31
Not recorded 7 19 2 13
27
The geometric means of one-week outdoor PM2.5 concentrations corresponding to indoor
residential PM2.5 measurement periods were 58.2 µg/m3 (95% CI: 52.9, 63.9 µg/m3) and
38.0 µg/m3 (95% CI: 34.2, 42.2 µg/m3) for the first and second measurements, respectively.
Outdoor concentrations were similar for control and intervention homes. Across the
seasons, geometric mean outdoor PM2.5 concentrations were 118.0 µg/m3 (95% CI: 110.4,
126.2 µg/m3) and 60.0 µg/m3 (95% CI: 54.9, 65.7 µg/m3) in winter and fall, and 31.7 µg/m3
(95% CI: 29.5, 34.0 µg/m3) and 20.3 µg/m3 (95% CI: 19.3, 21.3 µg/m3) in spring and
summer, respectively. One-week indoor and outdoor PM2.5 concentrations were correlated
(r=0.69, n=429), with higher correlations for control (r=0.78, n=203) than intervention
apartments (r=0.63, n=226; see Figure B.2).
The overall geometric means of one-week indoor PM2.5 concentrations were 22.5 µg/m3
(95% CI: 20.5, 24.6 µg/m3) and 18.3 µg/m3 (95% CI: 16.6, 20.1 µg/m3) for the first and
second measurements, respectively. Over half (64%) of the first home measurements were
made in fall and winter reflecting higher indoor PM2.5 concentrations compared with the
second measurements, the majority of which (61%) were made in spring and summer when
concentrations were lower. Overall, the intervention reduced indoor PM2.5 concentrations
by 29% (95% CI: 21, 37 %, Table 2). We observed larger reductions when the air cleaners
were first deployed in early pregnancy (40%, 95% CI: 31, 48%), compared with after
roughly five months of use (15%, 95% CI: 0, 27%, Table 2; Figure 3). Apartments that
received two air cleaners had larger reductions in PM2.5 concentrations (33%, 95 % CI: 25,
41%) than apartments with one air cleaner (20%, 95% CI: 6, 32%). This trend was seen for
measurements made both early and late in the air cleaners’ deployment. No differences in
effectiveness were observed for reported air cleaner use. Stratification by season revealed
a higher non-significant difference in air cleaner effectiveness between winter (36%, 95%:
20, 49%) and summer (18%, 95% CI: 4, 30%). Greater wintertime reductions were
observed for apartments where windows were opened less frequently. Significantly higher
indoor PM2.5 concentrations were seen in apartments of participants who lived with
smokers. Higher concentrations were also seen in apartments where smoking occurred
28
indoors, although differences were not significant. Behaviours related to smoking in the
home did not influence air cleaner effectiveness.
29
Table 2.2. Effect of portable HEPA filter air cleaner use on one-week residential indoor PM2.5 concentrations.
GM (95% CI) µg/m3 % changea (95% CI)
Control Intervention Crude Adjusted for outdoor
PM2.5
All data 24.5 (22.2, 27.0)
n = 212
17.3 (15.8, 18.8)
n = 235 -30 (-38, -22) -29 (-37, -21)
Duration of air cleaner use
First measurement
30.3 (26.7, 34.3)
n = 111
17.3 (15.4, 19.4)
n = 125 -43 (-52, -32) -40 (-48, -31)
Second measurement
19.4 (16.9, 22.3)
n = 101
17.3 (15.1, 19.7)
n = 110 -11 (-26, 8) -15 (-27, 0)
Number of air cleaners deployed
1 air cleaner -----
18.7 (15.7, 22.3)
n = 64 -23 (-35, -9) -20 (-32, -6)
2 air cleaners -----
16.7 (15.1, 18.4)
n = 167 -31 (-39, -21) -33 (-41, -25)
Air cleaner densityb
< 3.0 air cleaners/100 m2 -----
17.9 (15.7, 20.3)
n = 102 -27 (-36, -16) -28 (-37, -18)
≥ 3.0 air cleaners/100 m2 -----
16.5 (14.6, 18.6)
n = 123 -32 (-41, -22) -30 (-38, -20)
Air cleaner usec
< 63 % of study period -----
17.1 (14.7, 19.7)
n = 87 -30 (-40, -18) -33 (-42, -23)
≥ 63 % of study period ----- 17.2 (15.2, 19.4)
n = 121 -31 (-40, -20) -30 (-39, -20)
Season
Winter
44.5 (39.0, 50.9)
n = 59
28.5 (23.7, 34.4)
n = 54 -36 (-49, -20) -36 (-49, -20)
Spring
22.6 (19.3, 26.5)
n = 47
15.6 (13.6, 17.9)
n = 64 -31 (-44, -15) -35 (-48, -19)
30
Summer
11.7 (10.5, 13.1)
n = 53
9.5 (8.4, 10.8)
n = 51 -19 (-31, -4) -18 (-30, -4)
Fall
28.3 (23.9, 33.5)
n = 53
20 (17.5, 22.8)
n = 66 -29 (-43, -13) -31 (-43, -18)
Window opening in winter (Dec - Feb)
Open < half the month
46.9 (38.6, 57.1)
n = 34
25.7 (18.9, 35.0)
n = 29 -45 (-61, -22) -45 (-61, -22)
Open ≥ half the month
40.8 (34.0, 49.0)
n = 24
32.2 (26.4, 39.3)
n = 25 -21 (-39, 3) -23 (-4, 10)
Lived with a smoker at any time in pregnancy
No
22.8 (19.8, 26.3)
n = 98
16.0 (14.2, 18.2)
n = 120 -31 (-42, -19) -26 (-37, -13)
Yes
26.0 (22.7, 29.8)
n = 111
18.8 (16.6, 21.2)
n = 112 -28 (-39, -16) -29 (-40, -17)
Smoking occurred in the home at any time in pregnancy
No
24.2 (21.5, 27.2)
n = 144
16.7 (15.0, 18.6)
n = 155 -31 (-41, -19) -29 (-38, -19)
Yes
25.2 (21.2, 29.9)
n = 65
18.4 (16.0, 21.2)
n = 80 -32 (-44, -17) -33 (-45, -19)
aPercent reduction comparing one-week indoor PM2.5 concentrations in intervention to control apartments, except for analyses of number of air cleaners which
compares indoor PM2.5 concentrations in apartments with one and two air cleaners against apartments with no air cleaners. b3.0 air cleaners/100 m2 was the geometric mean air density calculated for intervention apartments. c63% was the geometric mean air cleaner use reported by participants.
31
Figure 2.3. Distribution of one-week indoor PM2.5 concentrations in control and
intervention homes stratified by season and measurement (the first
measurement was made when air cleaners were newly deployed and
the second measurement was made after approximately five months
of use).
The intervention reduced blood cadmium concentrations by 14% (95% CI: 4, 23%), from
a geometric mean of 0.23 µg/L (95% CI: 0.21, 0.25 µg/L) to 0.20 µg/L (95% CI: 0.19,
0.21 µg/L). The effect of the air cleaners on blood cadmium concentrations was not
significantly modified by smoking in the home, working outside the home, time spent at
home, or number of air cleaners.
Blood cadmium and hair nicotine concentrations were more strongly correlated among
participants who lived with a smoker (r=0.29, p=0.02, n=66) compared with those who did
not (r=0.10, p=0.47, n = 56). Blood cadmium concentrations were 14% (95% CI: 2, 28%)
higher among participants who lived with a smoker, and 24% (95% CI: 10, 41%) higher
32
among participants who lived in apartments where smoking occurred indoors, compared
with participants from non-smoking households. Similarly, geometric mean hair nicotine
concentrations were significantly higher among participants living in smoking (33 ng/mg;
95% CI: 0.23, 0.46 ng/mg) versus non-smoking households (0.10 ng/mg; 95% CI: 0.08.
0.14 ng/mg).
2.5. Discussion
In this relatively large randomized controlled trial we assessed the impact of HEPA filter
air cleaners on indoor PM2.5 and blood cadmium concentrations among pregnant women
in Ulaanbaatar, Mongolia. Air cleaners reduced one-week indoor PM2.5 concentrations by
29% (95% CI: 21, 37%) and blood cadmium concentrations by 14% (95% CI: 4, 23%).
Larger PM2.5 reductions were seen for the first measurement (40%, 95 % CI: 31, 48%),
when the air cleaners were newly deployed, compared with the second measurement (15%,
95% CI: 0, 27%), which was made after roughly five months of air cleaner use. We found
strong correlations between indoor and outdoor PM2.5, indicating that outdoor PM2.5
contributed substantially to indoor concentrations. Since filter effectiveness followed the
same seasonal pattern as outdoor PM2.5 concentrations, the impact of the intervention on
indoor PM2.5 concentrations was extraordinarily large in the winter months, when the
geometric mean was reduced from 45 to 29 µg/m3. Apartments with two air cleaners
experienced larger reductions in indoor PM2.5 than apartments with one air cleaner; in
contrast, we did not observe differences in HEPA cleaner effectiveness by the density of
air cleaners (number of air cleaners/100 m2). No differences in effectiveness were found
based on reported air cleaner use, which was crudely assessed from a question about overall
use and was not based specifically on the periods of air pollution monitoring.
The reductions in residential indoor PM2.5 in our study are consistent with findings reported
by other studies evaluating portable air cleaner use in residential settings.89-95,115,116 Only
one study has been conducted in a similarly highly polluted setting. Chen et al. (2015)
33
evaluated the use of portable electrostatic precipitator air cleaners in 10 university
dormitory rooms in Shanghai, China. The authors reported a 57% reduction in indoor
PM2.5, with mean (SD) concentrations decreasing from 96.2 (25.8) μg/m3 during a 48-hour
control period with sham filtration to 41.3 (17.6) μg/m3 during a 48-hour period with active
filtration.116 Four US-based randomized controlled trials evaluated the use of portable air
cleaners over six to 12 months in homes.90,92,93,115 Authors reported mean reductions in
PM2.5 or particle counts (> 0.3 µm) of 32-66% (see Table B.3). In the only study to assess
changes in air cleaner effectiveness over time, Lanphear et al. (2011) reported decreases in
particle count concentrations (> 0.3 µm) of 38% in intervention apartments compared with
control apartments after six months of air cleaner use, and a 25% reduction after 12 months
of use (n=225).93 In contrast, we saw greater decreases in effectiveness over the roughly
five months between air pollution measurements in our study. This larger decrease may
have been due to more rapid overloading of HEPA filters in this high pollution setting or
lower compliance to the intervention.
Overall, participants reported using air cleaners for 64% of the study period. Although we
did not systematically evaluate the reasons that participants shut off the air cleaners,
anecdotal reports from participants revealed concerns about noise and electricity costs. For
example, some participants reported consistently turning air cleaners off at night to
minimize noise. Studies measuring compliance to air cleaner use have previously reported
that participants used air cleaners approximately 34 to 79% of the time during study periods
ranging from one to 12 months.90,92,115,117 Batterman et al. (2012) also looked at changes in
compliance over time. The authors conducted one-week indoor air quality monitoring in
apartments for three to four consecutive seasons, with air cleaner use being monitored
throughout this period.90 Air cleaner use declined from a mean (SD) of 84% (24) during
the first indoor air quality measurement to 63% (33) when indoor air quality measurements
were collected in subsequent seasons. Compliance was lowest during periods outside of
when indoor air quality measurements were taken, with a mean use of 34% (30).90 Similar
to our study, Ward et al. (2017) reported no relationship between air cleaner effectiveness
34
and compliance, which was assessed by comparing expected and measured energy
consumption for air cleaner units during the study period.117
Air cleaners reduced average blood cadmium concentrations by 14%. A reduction in
cadmium exposures, even from low levels, could have important public health
implications.50 Cadmium is a known human carcinogen and has also been linked with
adverse cardiovascular and kidney effects.50,133 Among pregnant women, blood cadmium
concentrations have been linked to impaired fetal growth, as indicated by small for
gestational age14,62 and reduced birth weight.69 Tobacco smoke exposures have been
reported to be the greatest contributor to blood and urinary cadmium levels among
smokers.134 Similarly, among pregnant women, elevated blood cadmium levels have been
reported among those who were active smokers or exposed to SHS.135-137 Although blood
cadmium concentrations cannot definitively be linked to SHS exposures, we found three
pieces of compelling evidence to suggest that SHS exposure was an important source of
cadmium in our population. First, we found higher blood cadmium concentrations among
participants who reported living with smokers as well as among those living in homes
where smoking occurred indoors, compared with those in non-smoking households.
Second, we found higher correlations between blood cadmium and hair nicotine
concentrations among participants who lived with smokers (r=0.29) compared with those
who did not (r=0.10). Finally, we found lower blood cadmium concentrations among
intervention participants, suggesting that airborne exposures were lower in this group.
Other sources of airborne cadmium, including coal combustion, may have also contributed
to blood cadmium exposures.121 Our finding that air cleaner use decreased SHS exposure
differs from previous studies. Lanphear et al. (2011) and Butz et al. (2011) reported no
changes in hair, serum or urinary cotinine concentrations comparing intervention and
control participants, or pre- and post-intervention levels, among children using portable
HEPA filter air cleaners for six to 12 months.92,93
In our study, geometric mean blood cadmium and hair nicotine concentrations found
among participants with and without SHS exposures ranged from 0.20-0.23 µg/L and 0.10-
35
0.33 ng/mg, respectively, which are relatively low compared with previously reported
values among pregnant women. In a review of 24 studies assessing blood cadmium
concentrations among pregnant women, Taylor et al. (2014) reported mean and median
blood cadmium concentrations ranging from 0.09 to 2.26 µg/L among populations in
several countries, including Poland, Russia, South Africa, Egypt, India, Norway, France,
United States and China.138 Few guidelines or levels of concern exist for blood cadmium.
In Germany, a guideline of 1 µg/L has been established for the general public, which
includes non-smoking adults aged 18–69 years.138 Similarly, hair nicotine concentrations
in our study population were substantially lower than concentrations reported among
pregnant women living with partners who smoke (0.51 to 3.18 ng/mg).139,140
Our findings suggest that portable HEPA filter air cleaners are an effective household level
intervention to reduce PM2.5. The situation in Ulaanbaatar is similar to many other rapidly
growing cities, where already dramatically high pollution concentrations are expected to
increase, and strategies to effectively manage air quality will take years or decades to
implement.97 Proposed strategies in Ulaanbaatar have included dissemination of cleaner-
burning coal stoves and use of cleaner-burning fuels in ger households, as well as improved
emission controls for coal-fired power plants.97 Cigarette smoke is the most important
indoor source of PM2.5 in non-ger households in Ulaanbaatar.97 Nearly 40 % of Mongolian
men smoke,103 consistent with our finding that half of UGAAR participants lived with a
smoker and 34% lived with someone who smoked inside the home. Portable air cleaners
show promise because they are easy to operate and reduce concentrations inside residences,
where individuals spend the largest portion of time. The costs, which include an initial
purchase price typically starting at $200-300US,141,142 as well as maintenance and
operation costs, will be prohibitive for some. In addition, air cleaners must be appropriately
sized for the volume of the home and the air exchange rate and may not be a viable
intervention in situations when windows are frequently opened, or residences are not
tightly sealed. This is consistent with our finding that in winter months air cleaners were
more effective when windows were kept closed.
36
Some important limitations of our study should be noted. First, participants were not
blinded to the intervention. Previous air cleaner studies have used sham filtration to blind
participants to their intervention status, but instead of purchasing sham air cleaners we
chose to recruit a larger number of participants and deploy two air cleaners in larger
apartments. Moreover, our exposure measures were objective, which should minimize
potential bias resulting from the lack of blinding. Another limitation of our study is that we
did not replace HEPA filters during the study period, which has been done by others
assessing long-term air cleaner use and performance.90 We chose not to replace filters in
our study period to assess air cleaner efficacy under more “real world” conditions and to
minimize logistical challenges. Although we collected information on air cleaner use via
questionnaires and internal timers, data from timers were flawed and did not allow us to
assess how air cleaner use changed over time. We also did not assess air cleaner use in
gers, where the highest exposures in Ulaanbaatar occur and where exposure reduction is
needed most,97 due to concerns about the lack of reliable electricity and high air exchange
rates. Consequently, our findings on air cleaner effectiveness are likely not generalizable
to ger households. Finally, we approximated PM2.5 concentrations using the Dylos, a low-
cost optical particle counter. Extensive quality control and data cleaning steps identified
several instruments that provided unreliable data, which resulted in a large fraction of data
being removed prior to analysis. Despite these data losses, our analysis made use of an
extraordinarily large dataset (447 one-week indoor PM2.5 concentration measurements in
342 apartments). Consistent with several previous studies, we found excellent agreement
(R2=0.94) between Dylos particle counts and PM2.5 concentrations measured
gravimetrically.48,123-127
2.6. Conclusions
In this randomized controlled trial, we found that air cleaners substantially reduced indoor
PM2.5 concentrations and SHS exposures as measured by blood cadmium among a group
of pregnant women in a highly polluted city. Our findings suggest that portable air cleaners
37
are a useful household level intervention that can help reduce PM2.5 exposures during
pregnancy and other critical time periods.
38
Chapter 3.
The effect of portable HEPA filter air cleaner use during
pregnancy on fetal growth: the UGAAR randomized
controlled trial
3.1. Abstract
Background: Fine particulate matter (PM2.5) exposure may impair fetal growth.
Aims/Objectives: Our aim was to assess the effect of portable high efficiency particulate
air (HEPA) filter air cleaner use during pregnancy on fetal growth.
Methods: The Ulaanbaatar Gestation and Air Pollution Research (UGAAR) study is a
single-blind randomized controlled trial conducted in Ulaanbaatar, Mongolia. Non-
smoking pregnant women recruited at ≤ 18 weeks gestation were randomized to an
intervention (1-2 air cleaners in homes from early pregnancy until childbirth) or control
(no air cleaners) group. Participants were not blinded to their intervention status.
Demographic, health, and birth outcome data were obtained via questionnaires and clinic
records. We used unadjusted linear and logistic regression and time-to-event analysis to
evaluate the intervention. Our primary outcome was birth weight. Secondary outcomes
were gestational age-adjusted birth weight, birth length, head circumference, gestational
age at birth, and small for gestational age. The study is registered at ClinicalTrials.gov
(NCT01741051).
39
Results: We recruited 540 participants (272 control and 268 intervention) from January 9,
2014 to May 1, 2015. There were 465 live births and 28 losses to follow up. We previously
reported a 29% (95% CI: 21, 37%) reduction in indoor PM2.5 concentrations with portable
HEPA filter air cleaner use. The median (25th, 75th percentile) birth weights for control
and intervention participants were 3450 g (3150, 3800 g) and 3550 g (3200, 3800 g),
respectively (p=0.34). The intervention was not associated with birth weight (18 g; 95%
CI: -84, 120 g;), but in a pre-specified subgroup analysis of 429 term births the intervention
was associated with an 85 g (95% CI: 3, 167 g) increase in mean birth weight.
Conclusions: HEPA filter air cleaner use in a high pollution setting was associated with
greater birth weight only among babies born full term.
3.2. Introduction
Fine particulate (PM2.5) air pollution is a leading contributor to the global burden of disease
because exposure is ubiquitous and causes respiratory-, cardiovascular-, and cancer-related
morbidity and mortality.143 In 2016, 95% of the world’s population lived in areas where
PM2.5 concentrations exceeded the World Health Organization annual average guideline of
10 μg/m3.2 Although air pollution levels are decreasing in many high-income countries,
concentrations in low and middle-income (LMIC) countries remain unchanged or continue
to increase.105 Growing evidence from observational studies suggests that PM2.5 exposures
during pregnancy adversely affect fetal growth.4,7 A recent meta-analysis of 32 studies
presented pooled estimates of the effect of outdoor PM2.5 on birth weight and/or low birth
weight (LBW).4 A 10 µg/m3 increase in PM2.5 over the full duration of pregnancy was
associated with a 16 g (95% CI: 5, 27 g) reduction in birth weight and an increased risk of
LBW (odds ratio (OR)=1.09; 95% CI: 1.03, 1.15).4 Nearly all of the studies focused on
term births (≥37 weeks), and the majority were conducted in the US. The only randomized
controlled trial (RCT) of air pollution and fetal growth reported a trend toward greater birth
weight (89 g; 95% CI: -27, 204 g) among participants who used a chimney stove (n=69)
40
compared with those who used traditional open fires (n=105) during pregnancy in rural
Guatemala.144
Portable high efficiency particulate air (HEPA) filter air cleaners (henceforth “HEPA
cleaners”) are a promising intervention to lower PM2.5 exposures at the household level.
Their use has been shown to reduce indoor residential PM2.5 concentrations by 29-
65%.55,89,91,92,94,95 These reductions in concentrations can have large impacts on exposure
since individuals spend the majority of time indoors, and because HEPA cleaners target
both outdoor pollution that infiltrates indoors and indoor-generated pollution from
cigarettes, cooking, and other sources. The impact of portable air cleaners on fetal growth
has not been previously studied, but their short-term use (days to weeks) may induce
biological changes relevant to fetal growth,76 including improvements in endothelial
function,95 inflammation,91 and blood pressure.96 The objective of this randomized trial
was to determine if HEPA cleaner use at home from early pregnancy until childbirth among
pregnant women in Ulaanbaatar, Mongolia was associated with improvements in fetal
growth, compared with no HEPA cleaner use. Our primary motivation for this work was
to introduce an exposure gradient from which to investigate the causal role of PM2.5 on
fetal growth. Secondarily, we sought to evaluate HEPA cleaner use as a possible household
level intervention in high pollution settings.
3.3. Methods
Study design
The Ulaanbaatar Gestation and Air Pollution Research (UGAAR) study is a single-blind
RCT designed to assess the effect of portable HEPA cleaner use during pregnancy on fetal
growth and early childhood development in Ulaanbaatar, Mongolia (ClinicalTrials.gov:
NCT01741051). This city is among the world’s most polluted in wintertime, primarily due
to coal combustion for heating in low income neighborhoods and emissions from three coal
41
fired power plants.100 The population-weighted annual average PM2.5 concentration in
Ulaanbaatar is over seven times the World Health Organization (WHO) guideline
concentration of 10 μg/m3.100 Household coal use occurs in gers, traditional felt-lined yurts,
which house approximately 60% of the city’s population 98. Ulaanbaatar’s other residents
live in apartments that receive electricity and heat from the power plants. We previously
reported that HEPA cleaners reduced indoor PM2.5 by 29% (95% CI: 21, 37%) in this
apartment-dwelling study population, with larger reductions in homes that received two air
cleaners (33%, 95% CI: 25, 41%) than those that received one (20%, 95% CI: 6, 32%).55
The study was conducted at two branches of the Sukhbaatar district Health Center of
Ulaanbaatar. The study protocol was approved by the Simon Fraser University Office of
Research Ethics (2013s0016) and the Mongolian Ministry of Health Medical Ethics
Approval Committee (Decree No.7).
Participants
We recruited women who met the following criteria: ≥ 18 years of age, ≤ 18 weeks of a
single-gestation pregnancy, non-smoker, living in an apartment, planning to give birth in a
medical facility in the city, and not using a residential portable air cleaner at enrollment.
We excluded women who lived in ger households because electricity is unreliable in ger
neighborhoods and gers may have higher indoor-outdoor air exchange rates, which reduces
HEPA cleaner effectiveness. Moreover, gers generally have higher indoor pollution
emissions, and we were primarily interested in the effects of community air pollution. We
recruited participants at one of two reproductive health clinics in the centrally-located
Sukhbaatar district. This district was targeted due to its large population living in
apartments, its relatively high pollution concentrations, and our relationships with clinic
staff. All participants provided written informed consent prior to data collection.
Participants were compensated with 65,000 tugriks (approximately $30 USD) upon
completion of data collection; a pro-rated amount was provided to participants who
withdrew before completion of the study.
42
Randomization and blinding
We used simple randomization to assign participants to the intervention or control group
using sealed opaque envelopes containing randomly generated “filter” or “control”
allocations and labeled with participant identification numbers from one to 580 by a
principal investigator (RWA). Allocation was done on a 1:1 ratio. Once an individual was
deemed eligible and provided written consent, a sealed envelope was drawn in sequential
order and opened by a study coordinator who informed the participant of their allocation.
Only one envelope was opened per participant; if a participant did not agree to her
allocation she was not enrolled in the study. The envelope was then discarded and a new
one was opened when the next participant was enrolled. Participants were not blinded to
their intervention status.
Intervention
The intervention group received one or two HEPA cleaners (Coway AP-1009CH), based
on the size of their home, to use from enrollment to the end of pregnancy. The control
group received no air cleaners.
Procedures
We installed one HEPA cleaner in the main living area of apartments with areas < 40 m2;
in larger apartments a second HEPA cleaner was installed in the participant’s bedroom.
We installed air cleaners at the first home visit, which occurred shortly after enrollment,
and participants were encouraged to use the air cleaners continuously. The HEPA cleaners
have a clean air delivery rate for tobacco smoke (particles sized 0.09-1.0 µm) of 149 cubic
feet per minute, which is appropriate for rooms up to approximately 22 m2. Two features,
an internal PM sensor and light that changes colour based on PM concentration, were
disabled to avoid biasing participants’ behavior. The units were set to operate only on the
second-highest fan setting due to noise at the highest fan setting.
43
Data were collected at clinic visits that occurred shortly after enrollment (5-19 weeks
gestation) and later in pregnancy (24-37 weeks gestation; Figure 3.1). At these visits staff
collected data on housing, lifestyle, and maternal health via questionnaires. During the
second clinic visit we asked participants to estimate the percentage of time HEPA cleaners
were used. After birth, we obtained birth weight, length, head circumference, gestational
age, sex, and mode of delivery from clinic records. Participants self-reported the
occurrence and timing of spontaneous abortions, and information on stillbirths was
obtained from clinic records. We also collected information from clinic records on
pregnancy complications and co-morbidities, including pre-existing and gestational
diabetes and hypertension, anemia, and TORCH (Toxoplasmosis, Others [syphilis,
varicella-zoster, parvovirus B19], Rubella, Cytomegalovirus, Herpes) infections. A full
summary of UGAAR data collection activities is provided in the supplemental material
(Figure C.1).
Figure 3.1 Data collection
Outcomes
Our primary outcome was birth weight. We also analyzed secondary outcomes, including
gestational age-adjusted birth weight, birth length, head circumference, and small for
44
gestational age (SGA) at birth, as additional measures of intrauterine growth restriction, as
well as gestational age at birth. Birth weight was available to the nearest 10 g, and birth
length and head circumference were available to the nearest one cm. Gestational age was
available either as a completed week or as a one-week interval (e.g. 37-38 weeks); for the
latter, the mid-point of the interval was used (e.g. 37.5 weeks). SGA was defined as a birth
weight <10th percentile for sex and gestational age of the WHO fetal growth chart.145 We
also explored additional outcomes that were not pre-specified: ponderal index, LBW, and
preterm birth (PTB). Ponderal index was calculated as 100 multiplied by birth weight (g)
divided by crown-heel length cubed (cm3). Low birth weight was defined as < 2,500 g, and
PTB was defined as birth at <37 weeks gestation. Adverse events were spontaneous
abortion, stillbirth, and neonatal death. Spontaneous abortion and stillbirth were defined as
pregnancy loss at <20 weeks and ≥20 weeks, respectively. Neonatal death was defined as
a death occurring within 28 days of a live birth. Stillbirth weight was not included in the
analysis of birth weight since these data were not available.
Statistical analysis
Sample size calculations were based on term birth weight. We expected infants in the
intervention group to weigh 120 g more at birth, on average, than infants in the control
group (from a mean ± standard deviation of 3,490 g ± 520 g). This value was based on
previous estimates of outdoor PM2.5 effects on term birth weight,146,147 as well as
assumptions on indoor and outdoor PM2.5 concentrations,122 infiltration of outdoor
PM2.5,148 the effect of HEPA cleaners on indoor PM2.5,
91,94 and time spent in different
microenvironments during pregnancy.149 To detect a 120 g difference in mean birth weight
with a type I error rate of 0.05 (2-sided) and a type II error rate of 0.20, we estimated that
460 participants, in equal numbers in both treatment groups, were needed. We assumed
18% attrition due to dropout and pregnancy loss, so we targeted a population of 540
participants.
45
We used unadjusted linear and logistic regression to assess the effect of the intervention on
all continuous and categorical outcomes, respectively, except for gestational age. Since
gestational age had a non-normal distribution, we used time-to-event analysis to calculate
hazard ratios for time to a live birth comparing intervention with control participants; we
censored lost to follow up participants and treated pregnancy loss from spontaneous
abortion or stillbirth as a competing risk. Models of gestational age-adjusted birth weight
were adjusted for gestational age using linear and quadratic terms to account for the non-
linear relationship between fetal growth and gestational age. We conducted a complete case
analysis among all live births, excluding those involving a chromosomal abnormality, as
well as participants who were lost to follow up or who had a pregnancy loss. Participants
were analyzed according to their original intervention assignments, regardless of which
treatment they were given or used.
Because nearly all of the existing observational evidence,4 including the studies used in our
sample size calculations, is based on term births we stratified the analyses by gestational
age (all births and term births, defined as ≥ 37 weeks) in a priori planned analyses. We
also tested effect modification using stratified analyses and interaction terms in the models
for variables identified a priori: gestational age at birth, exposure to second hand smoke
(SHS), average time spent indoors at home during pregnancy, sex of the baby, season of
birth, and self-reported air cleaner use. We additionally investigated income, as a proxy for
socioeconomic status, as a potential effect modifier post hoc. Finally, we repeated all
analyses to also estimate the effect of one and two air cleaners on our outcomes.
We assessed the sensitivity of the intervention effect estimates for birth weight to different
factors. First, as an alternative to our intention-to-treat analysis, we estimated the effect of
the intervention based on the treatment that participants received (i.e “per protocol”). We
also estimated intervention effects on birth weight after excluding (i) neonatal deaths, (ii)
participants who reported smoking at any time in pregnancy, and (iii) potential errors in
gestational age or birth weight (identified as observations that exceeded the WHO fetal
46
growth chart 95th percentiles of birth weight for gestational age and sex by ≥20%). Finally,
we estimated effects while adjusting for anemia status and PTB in the regression models.
Role of funding source
This study was funded by the Canadian Institutes of Health Research. Woongjin-Coway
provided modified and discounted air cleaners. The funder and the company had no role in
study design, data collection, data analysis, interpretation of study findings, or manuscript
preparation.
3.4. Results
We recruited 540 participants from January 9, 2014 to May 1, 2015. Participants were
enrolled in the study at a median (25th, 75th percentile) gestation of 10 weeks (8,
12 weeks). Two hundred and seventy-two participants were randomized to the control
group and 268 were randomized to the intervention group (Figure 3.2). Five participants
that were allocated to the control group mistakenly received the intervention, while three
participants allocated to the intervention group did not receive HEPA cleaners; data from
these participants were analyzed according to their original treatment assignments.
Twenty-eight (5%) participants were lost to follow up, and there were 34 (6%) spontaneous
abortions and 13 stillbirths (2%). In total, 465 participants (86%) had a known live birth,
five of which resulted in neonatal deaths (1%) due to heart defects (n=2), respiratory failure
(n=2), and brain injury (n=1). Two children were excluded due to chromosomal
abnormalities (trisomy 18 and 21), leaving 463 live births in our complete case analysis.
47
Figure 3.2 Trial profile
Participants lost to follow up and those who remained in the study had similar
demographics and lifestyles, including age at enrollment, household income, maternal
education, marital status, parity, and pre-pregnancy BMI (Table C.1). Some participants
withdrew consent before our first questionnaire was administered resulting in more missing
observations for these variables. More control participants were lost to follow up (19 vs. 9;
p=0.06).
Baseline characteristics were similar among control and intervention participants (Table
3.1). The median (25th, 75th percentile) age at enrollment was 29 years (25, 33 years).
Control and intervention participants were enrolled at median (25th, 75th percentile)
48
gestational ages of 11 (9, 12) and 11 (9, 13) weeks, respectively. The seasonal pattern of
enrollment was also similar between the groups, with the most participants enrolled in
winter (32%) and the fewest participants enrolled in summer (12%). Over half of the
participants in both groups reported a monthly household income at or above the city’s
average of 800,000 Tugriks (approximately $360 US).150 Most participants (≥80%)
completed university or college and reported being married/common-law. Although all
participants identified as non-smokers at enrollment, 8% of participants in both groups
reported smoking in early pregnancy. Parity was similar between groups. Control
participants had a shorter interval from the last pregnancy (24 vs 31 months), although the
response rate to this question was poor with over 40% of observations missing. The
intervention was implemented at a median (25th, 75th percentile) gestation of 11 weeks (9,
13 weeks). Seventy households received one HEPA cleaner and 186 households received
two HEPA cleaners. The ratio of air cleaners to apartment area was similar for participants
who received one (median: 3.0 air cleaners/100 m2; 25th, 75th percentile: 2.3, 3.7 air
cleaners/100 m2) and two (median: 3.4 air cleaners/100 m2; 25th, 75th percentile: 2.4, 3.9
air cleaners/100 m2) HEPA cleaners. Participants reported using HEPA cleaners for a
median of 70% of the time; reported use did not differ by the number of air cleaners
received. We previously reported that outdoor PM2.5 concentrations measured at centrally-
located government monitoring stations during the study period were similar for control
and intervention participants.55
49
Table 3.1. Summary of baseline characteristics for control and intervention
participants.
Control
(n = 223)
Intervention
(n = 240)
Median (25th, 75th
percentile)
or N (%)
Median (25th, 75th
percentile)
or N (%)
Mother’s age at enrollment, years 28 (25, 33) 30 (25, 33)
Gestational age at enrollment, weeks 11 (9, 12) 11 (9, 13)
Season of enrollment
Winter (December, January, February) 77 (34) 71 (30)
Spring (March, April, May) 66 (30) 63 (26)
Summer (June, July, August) 23 (10) 33 (14)
Fall (September, October, November) 57 (26) 73 (30)
Monthly household income
< 800,000 Tugriksa 69 (31) 83 (34)
≥ 800, 000 Tugriks 150 (67) 155 (65)
Not reported, N (%) 4 (2) 2 (1)
Mother’s education
Completed university 179 (80) 191 (80)
Did not complete university 29 (13) 28 (11)
Not reported, N (%) 15 (7) 21 (9)
Marital status
Married/common-law 184 (83) 207 (86)
Not married/common-law 39 (17) 33 (14)
Not reported, N (%) 0 (0) 0 (0)
Worked/volunteered outside the home
No 69 (31) 78 (32)
Yes 151 (68) 160 (67)
Not reported, N (%) 3 (1) 2 (1)
Parity
0 21 (9) 24 (10)
1 88 (40) 86 (36)
≥2 44 (20) 59 (24)
Not reported, N (%) 70 (31) 71 (30)
Time since last pregnancy, monthsb 24 (10, 51) 31 (15, 61)
Not reported, N (%) 99 (44) 105 (44)
Previous poor pregnancy outcomec
No 42 (19) 54 (23)
Yes 50 (22) 54 (23)
Not reported, N (%) 131 (59) 132 (54)
Pre-pregnancy BMI, kg/m2 21.7 (19.6, 23.9) 21.4 (19.8, 24.0)
Not reported, N (%) 21 (9) 8 (3)
Time spent indoors at home in early
pregnancy, hr/day 16.0 (14.0,18.7) 16.1 (14.0, 19.0)
Not reported, N (%) 46 (21) 63 (26) aApproximate average monthly income in Ulaanbaatar in 2014.150 At the time of data collection, 800,000
Tugriks was the equivalent of approximately $360 US.
50
bDefined as the period between the end of the last pregnancy, including live births and pregnancy losses, and
start of current pregnancy. cPrevious poor outcome included spontaneous abortion, still birth, low birth weight, macrosomia, ectopic
pregnancy, birth defect, and intrauterine growth restriction.
Maternal weight gain, second hand smoke exposure, and health complications during
pregnancy were similar between groups (Table 3.2). Roughly half of control and
intervention participants reported living with a smoker. A higher frequency of intervention
participants reported anemia during pregnancy (22% vs 15%; p=0.07). No control or
intervention participants had diabetes or gestational diabetes. Similarly, few participants
had hypertension, gestational hypertension, or TORCH infections.
51
Table 3.2. Summary of variables assessed during pregnancy
Control
(n = 223)
Intervention
(n = 240) p-
valuea
Median (25th, 75th
percentile)
or N (%)
Median (25th, 75th
percentile)
or N (%)
Weight gain during pregnancy
(kg)b 12 (8, 15) 11 (8, 15)
0.93
Not reported 43 (19) 31 (14)
Smoked at any time in pregnancy
No 203 (91) 217 (91) 0.99
Yes 19 (9) 20 (8)
Not reported, N (%) 1 (0) 3 (1)
Lived with a smoker at any time in pregnancy
No 106 (48) 121 (51)
0.64 Yes 112 (50) 115 (48)
Not reported, N (%) 5 (2) 4 (1)
Health during pregnancy
Anemia 34 (15) 53 (22) 0.07
Not reported, N (%) 0 (0) 0 (0)
Hypertension 11 (5) 13 (5) 0.84
Not reported, N (%) 0 (0) 0 (0)
Gestational hypertension 16 (7) 16 (7) 0.85
Not reported, N (%) 23 (10) 15 (6)
TORCH infectionsc 3 (1) 5 (2) 0.72
Not reported, N (%) 10 (4) 4 (2)
Type of delivery
Caesarean delivery 88 (39) 86 (36) 0.50
Vaginal delivery 135 (61) 154 (63)
Sex of child
Female 108 (48) 109 (45) 0.58
Male 115 (52) 131 (55)
Season of birth
Winter (Dec, Jan, Feb) 26 (12) 35 (15)
0.56 Spring (Mar, Apr, May) 52 (23) 59 (24)
Summer (Jun, Jul, Aug) 70 (31) 78 (33)
Fall (Sep, Oct, Nov) 75 (34) 68 (28) ap-values were generated using Fisher’s exact tests, 2-sample t-tests, and Mann-Whitney tests as appropriate.
bFrom approximately week 11 to week 31. cToxoplasmosis, Others [syphilis, varicella-zoster, parvovirus B19], Rubella, Cytomegalovirus, Herpes
infections.
The birth weight distributions were skewed by PTBs (Figure 3.3). The median (25th, 75th
percentile) birth weights for control and intervention participants were 3450 g (3150,
3800 g) and 3550 g (3200, 3800 g), respectively (p=0.34; Table 3.3). Regression results
52
indicated no significant intervention effect on mean birth weight (18 g; 95% CI: -84, 120
g), but after adjusting for PTB the intervention was associated with an 84 g (95% CI: -1,
170 g) increase in birth weight (Table C.2). The effect estimates from our other sensitivity
analyses were generally similar to those from the main analysis (Table C.2). Gestational
age, birth length, head circumference, ponderal index, and frequency of LBW and SGA
were similar for control and intervention participants (Table 3.3). The intervention was
associated with a significantly elevated risk of PTB (10% vs. 4%; OR=2.37; 1.11, 5.07).
When stratified by late (34-36 weeks) versus early (<34 weeks) cases, a significantly
elevated risk remained for late PTB (OR=3.55; 95% CI: 1.29, 9.73) but not early PTB
(OR=1.18; 95% CI: 0.36, 3.94).
53
Figure 3.3 Distribution of birth weight by treatment assignment for all births (a) and term births (b).
54
Table 3.3. Effect of the intervention on fetal growth and birth outcomes
Outcome
Median (25th, 75th percentile) or N (%) Effect of intervention
Control
n = 223
Intervention
n = 240
p-valuea Measure of association
All births
n = 463
Term birthsb
n = 429
Birth weight, g 3450
(3150, 3800)
3550
(3200, 3800) 0.34
Mean difference
(95% CI)
18
(-84, 120)
85
(3, 167)
Gestational age-adjusted
birth weight, gc --- --- ---
Mean difference
(95% CI)
48
(-31, 126)
81
(2, 159)
Birth length, cm 50
(50, 52)
51
(50, 52) 0.63
Mean difference
(95% CI)
-0.01
(-0.53, 0.51)
0.32
(-0.04, 0.68)
Head circumference, cm 35 (34, 36) 35
(34, 36) 0.23
Mean difference
(95% CI)
-0.07
(-0.4, 0.26)
0.14
(-0.13, 0.4)
Ponderal index, g/cm3 2.6
(2.5, 2.8)
2.6
(2.5, 2.8) 0.92
Mean difference
(95% CI)
0.01
(-0.04, 0.07)
0.02
(-0.02, 0.06)
Gestational age 39.5
(38.5, 40.0)
39.5
(38.5, 40.0) 0.87
Hazard ratio for
time to a live birth
(95% CI)
1.12
(0.96, 1.32)
1.06
(0.90,1.25)
Small for gestational age 18 (8) 16 (7) 0.67 Odds ratio (95% CI) 0.81
(0.4, 1.64)
0.44
(0.19, 1.05)
Low birth weightd 10 (4) 13 (5) 0.60 Odds ratio (95% CI) 1.22
(0.52, 2.84) -----
Preterm birth 10 (4) 24 (10) 0.03 Odds ratio (95% CI) 2.37
(1.11, 5.07) -----
ap-values were generated using non-parametric Wilcoxon rank tests for continuous outcomes and Fisher’s exact tests for categorical outcomes. bBirths occurring ≥ 37 weeks gestation. cModels were adjusted for gestational week and gestational week squared.
dThere were no cases of low birth weight among term births in the intervention group.
55
Among term births, the median (25th, 75th percentile) birth weights for control and
intervention participants were 3500 g (3200, 3800 g) and 3600 g (3300, 3850 g),
respectively. The intervention was associated with greater term birth weight (85 g; 95%
CI: 3, 167 g; Figure 3.4) and gestational age-adjusted term birth weight (81 g; 95% CI: 2,
159 g), as well as a trend toward decreased risk of SGA (OR=0.44; 95% CI: 0.19, 1.05).
No variables modified the intervention effects for all births or term births (Figure 3.4).
56
Figure 3.4. Effect of the intervention on birth weight in stratified analyses for (a) all births and (b) term births. Note: 16 hours per day was the median time spent at home and 70% was the median self-reported use of the HEPA air cleaners. Numbers (Ns) reported for self-
reported HEPA cleaner use reflect the number of intervention homes and effect estimates are relative to control homes. Monthly household income, as a proxy
for socioeconomic status, was investigated as a potential effect modifier post hoc.
57
We also estimated the effects of using one or two air cleaners on birth outcomes separately because
we previously reported greater PM2.5 reductions among participants who received two air
cleaners55 Use of one HEPA cleaner was not associated with significant differences in birth weight
(26 g; 95% CI: -100, 151 g) or gestational age-adjusted birth weight (17 g; 95% CI: -104, 138 g)
among term births. For participants who received two HEPA cleaners, the intervention was
associated with non-significant trends toward greater birth weight (85 g; 95% CI: -7, 177 g) and
gestational age-adjusted birth weight (82 g; 95% CI: -5, 168 g) among term births.
There were 47 adverse events not prespecified as secondary outcomes, including 34 spontaneous
abortions (24 control, 10 intervention) and 13 stillbirths (5 control, 8 intervention). Participants
who had a spontaneous abortion were enrolled at a median (25th, 75th percentile) gestation of 8.0
weeks (7.0, 9.0 weeks), and participants who had a stillbirth were enrolled at a median (25th, 75th
percentile) gestation of 10.0 weeks (5.0, 12.0 weeks). Spontaneous abortions occurred at a median
(25th, 75th percentile) gestation of 12.0 weeks (9.5, 14.0 weeks) and stillbirths occurred at a
median (25th, 75th percentile) gestation of 30.0 weeks (20.5, 36.0 weeks). There were no
significant differences in the timing of spontaneous abortions or stillbirths between groups. The
intervention was associated with a decreased risk of spontaneous abortion (OR=0.38; 95% CI:
0.18, 0.82), but there was no association with stillbirth (OR=1.58; 95% CI: 0.51, 4.90).
3.5. Discussion
In this single-blind RCT in a community exposed to very high air pollution concentrations,
portable HEPA cleaner use during pregnancy was not associated with improvements in fetal
growth among all births. However, the intervention was associated with an 85 g (95% CI: 3, 167 g)
increase in term birth weight. Unexpectedly, we also saw an increased risk of PTB (OR=2.37; 95%
CI: 1.11, 5.07) and a decreased risk of spontaneous abortion (OR=0.38; 95% CI: 0.18, 0.82) with
intervention use.
58
Fetal growth restriction and shorter duration of gestation can both cause reductions in birth weight.
To reduce the influence of gestational age and evaluate the effect of air pollution on “normal” fetal
growth, others have restricted their investigations to term births.4,28,151 We observed a 100 g greater
median birth weight among the intervention group in our full study population, but this difference
was not statistically significant based on a non-parametric test of medians or our linear regression
model that tested for differences in means. However, adjusting for PTB resulted in a nearly
identical intervention effect estimate (84 g; 95% CI: -1, 170 g) as that of our subgroup analysis of
term births (85 g; 95% CI: 3, 167 g). This suggests that the intervention improved fetal growth in
the full study population, but that improvement was offset by a higher frequency of PTB in the
intervention group.
The effects of air pollution on fetal growth may be most detrimental in later pregnancy. For
example, in their recent meta-analysis, Sun et al. (2016) suggested that the second and third
trimesters might be a critical exposure window for PM2.5.4 Similarly, Rich et al. (2015) reported
that infants born to women in Beijing who had their eighth month of pregnancy during the 2008
Olympics, when outdoor pollution levels were substantially reduced, had a 23 g (95% CI; 5, 40 g)
greater mean term birth weight compared with infants whose eighth month fell in the same time
period in the year before or after the Olympics.152 Studies using repeated ultrasound measurements
also suggest that air pollution may not alter growth trajectories until late in the second
trimester.153,154
The observed relationship between the intervention and birth weight is biologically plausible. Air
pollution has been linked with changes relevant to fetal growth, such as systemic oxidative stress,
inflammation, endothelial dysfunction, blood coagulation, and blood pressure.155 During
pregnancy, these mechanisms are thought to decrease utero-placental blood flow from improper
placental vascularization and increase blood flow resistance, among other effects.76 The
pathophysiology of growth restriction is thought to be rooted in the inability of the fetus to receive
adequate nutrients and oxygen due to placental dysfunction brought upon by these mechanisms.27
Constituents adsorbed onto particles may act through additional pathways. For example, heavy
59
metals such as cadmium may deregulate processes such as calcium transport and placental release
of progesterone.83 Randomized trials of portable air cleaner use over 4-14 days have demonstrated
improved endothelial function, and reduced systemic inflammation and blood pressure, in healthy
adults (20 to 75 years of age).91,95,96 A recent RCT also reported lower diastolic blood pressure
among 162 pregnant women using cleaner burning ethanol stoves compared with 162 pregnant
women using more polluting kerosene or firewood stoves in Ibadan, Nigeria (p=0.04).156
The magnitude of our PTB-adjusted birth weight result and our term birth weight result is
comparable with the only previous randomized trial of air pollution and fetal growth. Thompson
et al. (2011) randomized 69 pregnant women in rural Guatemala to receive a chimney stove, and
105 pregnant women to a control group that continued to use open fires for cooking.144 The
intervention was associated with a 39% reduction in carbon monoxide exposure and an 89 g (95%
CI: -27, 204 g) increase in mean birth weight.144 Our estimated intervention effects on fetal growth
are also comparable to those from maternal nutrition interventions aimed at increasing birth
weight. A pooled analyses of 19 randomized trials conducted in countries such as Taiwan, India,
Iran, England, and the US reported 74 g (95% CI: 30, 117 g) greater mean birth weights among
women receiving protein energy supplementation compared with control participants consuming
routine diets.157 Another pooled analyses reported that dietary interventions were associated with
greater mean birth weights of 94 g, starting from a weight of 3086 g, in low-income countries and
49 g, from a starting weight of 3406 g, in high-income countries.158 Maternal malnutrition is a key
contributor to poor fetal growth, and like air pollution, it disproportionately affects populations in
LMIC.3,159 However, unlike air pollution-related interventions, considerable emphasis has been
placed on maternal nutrition intervention programs in LMIC to improve pregnancy and birth
outcomes.160
We unexpectedly found the intervention was associated with a decreased risk of spontaneous
abortion. There was no difference in timing of enrollment or spontaneous abortion between
intervention and control participants. Active smoking and SHS exposures during pregnancy have
been linked to spontaneous abortion,161 but the evidence for air pollution is conflicting. Enkhmaa
60
et al. (2014) reported strong positive correlations between monthly outdoor air pollution
concentrations and monthly hospital admissions for spontaneous abortion in Ulaanbaatar (r
>0.80).162 Monthly rates of spontaneous abortion were approximately 2.5 times greater in winter
than in summer. The investigators did not consider other seasonally varying factors such as vitamin
D or influenza exposure. In contrast, other studies conducted in Brazil163 and China164 reported no
associations between outdoor PM and spontaneous abortion.
Although we also found an increased risk of PTB among intervention participants, the intervention
did not significantly impact the risk of early (<34 weeks) PTBs (OR=1.18; 95% CI: 0.36, 3.94),
which are less likely to reflect iatrogenic intervention. The risk of late (34-36 weeks) PTBs was
significantly elevated among intervention participants. The reason for the surprising increase in
PTB in the intervention group may be found, in part, in the higher frequency of spontaneous
abortions in the control group; the presence of the intervention may have enabled fetuses to survive
long enough to be born preterm. The downstream effects of PM2.5-induced oxidative stress and
inflammation on fetal growth may be seen in early pregnancy, resulting in pregnancy loss, or in
later pregnancy, resulting in PTB and/or SGA.165-168 Although unlikely, an increased risk caused
by noise from the HEPA cleaners cannot be ruled out. No studies have investigated the potential
effects of air cleaner-related noise, but limited research suggests that exposure to noise from
aircraft and road traffic during pregnancy may increase the risk of PTB, possibly by increasing the
release of stress hormones that interfere with the production and release of progesterone.169 Our
findings of an adverse effect on PTB and a decreased risk of spontaneous abortion merit further
investigation.
Some important limitations of this study should be noted. Participants were not blinded to their
intervention status, which likely contributed to the greater loss to follow-up among control
participants. Although the treatment groups were similar in age, marital status, household income,
pre-pregnancy BMI, and parity, it is possible that the groups differed in unmeasured ways despite
randomization. Participants received air cleaners at a median of 11 weeks gestation so the HEPA
cleaners did not influence exposure during much of the first trimester. Spontaneous abortions were
61
based on participant report, and we were not able to distinguish between spontaneous and
medically indicated cases of PTB. Gestational age at birth was assessed from clinic records and
was based on a combination of first trimester ultrasound, last menstrual period, and clinical
assessment (symphyseal-fundal height measurements and/or Dubowitz or Ballard score).
Participants reported using the HEPA cleaners for a median 70% of the study period but we were
unable to assess how/if use changed throughout the study period. Anecdotal reports from
participants indicated that HEPA cleaner use may have been reduced due to concerns about noise
and electricity costs, and this may have reduced the benefits of the intervention.
3.6. Conclusions
Our motivations for studying the impact of HEPA cleaners on fetal growth were to investigate the
causal role of PM2.5 on fetal growth and to assess this household level intervention. We previously
reported that HEPA filter air cleaner use was associated with significant reductions in indoor
residential PM2.5 concentrations.55 In the present study, HEPA cleaner use was associated with
greater birth weight, but the effect was offset by a higher frequency of PTB in the intervention
group. We speculate that the apparent increase in risk of PTB was due, at least in part, to selection
bias resulting from a reduction in spontaneous abortions. Our findings provide additional evidence
for the health benefits of reducing air pollution. While HEPA cleaners can reduce exposures at the
household level, this intervention is not accessible to or appropriate for everyone. Portable air
cleaners require a constant supply of electricity, have costs related to initial purchase, operation,
and maintenance that may be prohibitive to some, and are generally less effective in dwellings
with high air exchange rates such as temporary or poorly constructed structures or in warm
climates where windows are frequently opened. Thus, in any community, relying solely on such
household level interventions to address air pollution exposures will not protect everyone,
particularly those most vulnerable. In the long-term, strategies to reduce community-wide air
pollution concentrations are needed to ensure that the benefits of exposure reduction are available
to all.
62
Chapter 4.
Gestational cadmium exposure and fetal growth in Ulaanbaatar,
Mongolia
4.1. Abstract
Background: Gestational cadmium exposure may impair fetal growth. Coal smoke has largely been
unexplored as a source of cadmium.
Aims/Objective: We investigated the relationship between gestational cadmium exposure and fetal
growth and assessed coal smoke as a potential source of cadmium among non-smoking pregnant
women in Ulaanbaatar, Mongolia, where residential coal combustion is a major source of air
pollution.
Methods: This observational study was nested within the Ulaanbaatar Gestation and Air Pollution
Research (UGAAR) study, a randomized controlled trial of portable high efficiency particulate air
(HEPA) filter air cleaner use during pregnancy, fetal growth, and early childhood development.
We measured third trimester blood cadmium concentrations in 374 out of 465 participants who
had a live birth. We used multiple linear and logistic regression to assess the relationships between
log2-transformed maternal blood cadmium concentrations and birth weight, length, head
circumference, ponderal index, low birth weight, small for gestational age, and preterm birth in
crude and adjusted models. We also evaluated the relationships between log2-transformed blood
cadmium concentrations and the density of coal-burning stoves within 5,000 m of each
participant’s apartment.
63
Results: The median (25th,75th percentile) blood cadmium concentration was 0.20 µg/L (0.15,
0.29 µg/L). A doubling of blood cadmium was associated with an 86 g (95% CI: 26, 145 g)
reduction in birth weight in adjusted models. An interquartile range increase in coal stove density
surrounding participants’ homes was associated with an 11% (95% CI: 0, 23%) increase in blood
cadmium concentrations.
Conclusions: Gestational cadmium exposure was associated with reduced birth weight. In settings
where coal is a widely used fuel, cadmium may play a role in the putative association between air
pollution and impaired fetal growth.
4.2. Introduction
Cadmium is a ubiquitous metal that is linked to cancer and kidney and bone disease.50,51 Diet and
smoking are the most important sources of cadmium.50 Foods with the highest cadmium content
include those grown or harvested in cadmium-rich environments such as leafy vegetables, grains,
shellfish, and organ meats.50 Tobacco smoke is the largest source of cadmium among smokers,51
while second hand smoke (SHS) is an important source among non-smokers.52-55 Cadmium can
also be emitted into the environment from the combustion of solid fuel, such as coal, and through
iron and copper smelting, waste incineration, and battery manufacturing and recycling.51,56,170,171
Gestational cadmium exposure may impair fetal growth. Several studies have linked maternal
cadmium exposures to decreases in birth weight11-13,15,17,20,60,63,69 and increased risks of low birth
weight (LBW),22,64 small for gestational age (SGA),14,23,59,62 and preterm birth (PTB),71 while
others have reported no associations.16,17,21,24,66,68,70 The inconsistent findings are likely due, in
part, to differences in study design, sample size, and exposure levels. Most studies have focused
on tobacco smoke or diet as the main sources of exposure. Few studies have been conducted among
non-smoking pregnant populations in communities heavily impacted by coal smoke, with these
studies reporting mixed findings on the importance of coal as a cadmium source.17,74,75
64
This study is part of the Ulaanbaatar Gestation and Air Pollution Research (UGAAR) study, a
randomized controlled trial of portable high efficiency particulate air (HEPA) filter air cleaner use,
fetal growth, and early childhood development (ClinicalTrials.gov Identifier: NCT01741051).
Ulaanbaatar, the capital of Mongolia, has some of the world’s highest air pollution concentrations.
The population-weighted annual average PM2.5 concentration is over seven times the World Health
Organization (WHO) guideline concentration of 10 μg/m3.100 The high pollution levels are
primarily due to wintertime residential coal use in neighborhoods of traditional Mongolian felt-
lined yurts (gers) and poorly constructed one or two-story wood and brick homes. Roughly 60%
of the city’s residents live in these neighborhoods, which surround the apartment-dwelling
population in the city center from which the UGAAR population was recruited.99 Three coal-fired
power plants supply electricity and heat to apartments and other buildings in the city.98 The plants
are a relatively minor contributor to PM2.5 concentrations and spatial variability compared with
residential coal emissions, which are responsible for 45-70% of total PM2.5 concentrations in the
city.98-100
We previously reported that SHS exposure was a source of cadmium based on our assessment of
third trimester blood cadmium concentrations among this study population.55 Close to half of our
non-smoking study participants reported living with a smoker during pregnancy, which was not
surprising considering the high smoking rates in Mongolia; nearly 40% of men smoke compared
with roughly 7% of women.100,103 We also assessed hair nicotine concentrations, as a marker of
SHS, among a subset of the population (n=125). We found higher blood cadmium and hair nicotine
concentrations, and stronger correlations between these measures, among participants who lived
with a smoker (r=0.29 vs r=0.10; n=125).55 We also reported that among this population, the HEPA
cleaner intervention was associated with a 14% (95% CI: 4, 23%) reduction in blood cadmium
concentrations,55 suggesting that airborne cadmium contributed to exposure, and an 85 g (95% CI:
3, 167 g) increase in term birth weight (see Chapter 3). Here, we investigated the relationship
between maternal blood cadmium concentrations and fetal growth. We also assessed residential
coal stoves as a source of gestational cadmium exposure.
65
4.3. Methods
Data collection
The UGAAR study has been described previously.55 Briefly, we enrolled women who were ≥ 18
years, ≤ 18 weeks into a single-gestation pregnancy, non-smokers, living in an apartment (i.e., not
living in a ger neighborhood), planning to give birth in an Ulaanbaatar maternity hospital, and not
using a portable air cleaner in the home at enrollment. We excluded women who lived in ger
households because electricity is unreliable in these neighbourhoods and because we wanted to
minimize the influence of indoor pollution emissions from coal stoves since we were primarily
interested in the effects of community air pollution. Recruitment was conducted at two
reproductive health clinics in the city’s centrally-located Sukhbaatar district. Participants were
randomly assigned to the intervention or control group. The intervention group received one or
two HEPA filter air cleaners (henceforth, “HEPA cleaners”), depending on the size of their
apartment, to use from early pregnancy until delivery, while the control group received no HEPA
cleaners. Data were collected from medical records and at two clinic visits, which occurred in early
(median of 11 weeks) and later pregnancy (median of 31 weeks).
Blood cadmium concentrations
Whole blood samples were collected by a nurse at the reproductive health clinic during the second
clinic visit. In total, 378 samples were collected from 465 participants who went on to have a live
birth. Some participants refused or were unavailable to provide a blood sample. Samples were
refrigerated and shipped to the Wadsworth Center (New York State Department of Health, Albany,
New York, USA) and analyzed within six weeks of collection using quadrupole-based inductively
coupled plasma-mass spectrometry (ICP-MS), with matrix-matched calibration.129 The limit of
quantification (LOQ) was 0.043 µg/L. Two samples that were below the LOQ were assigned
values of LOQ/2.130
66
Fetal growth outcomes
Our outcomes were birth weight, length, head circumference, ponderal index, LBW, SGA, and
PTB. Data on birth weight, birth length, head circumference, and gestational age were obtained
from clinic records after each delivery. We calculated ponderal index, a ratio of height to weight,
as 100 multiplied by birth weight (g) divided by crown-heel length cubed (cm3). Low birth weight
was defined as <2,500 g, SGA was defined as <10th percentile for gestational age and sex using
World Health Organization fetal growth charts145, and PTB was defined as birth at <37 weeks
gestation.
Determinants of cadmium exposure and co-variates
We collected information on potential sources of exposure as well as demographic and health
factors via questionnaires that were administered at two clinic visits. Exposure to tobacco smoke
was self-reported; we asked participants if they smoked and whether they lived with a smoker
during pregnancy. Previously, we validated the self-reported data and found significantly higher
median (25th, 75th percentile) hair nicotine concentrations [0.23 ng/mg (0.14, 0.72 ng/mg), n=66]
among participants that reported living with a smoker in late pregnancy compared with participants
who did not [0.09 ng/mg (0.05, 0.17 ng/mg), n=56, p < 0.001].55 Here, we used the self-reported
measures of tobacco smoke exposure since we had more complete data on these measures than on
hair nicotine. We used responses collected in late pregnancy to reflect the relevant exposure period
captured by concentrations in blood, which is roughly three to four months.50 We used coal stove
density as a proxy of residential coal smoke. We previously used high resolution aerial imagery
and object-based image classification to map the location of over 108,000 gers in Ulaanbaatar,172
and here we used those ger locations to approximate the density of coal-stoves surrounding each
participant’s apartment within a 5,000 m redius. The density of gers within 5,000 m of participants’
apartments was predictive of indoor PM2.5 concentrations in the UGAAR cohort and the locations
of ger neighbourhoods explained 66% of the spatial variability in outdoor sulphur dioxide (SO2)
concentrations, a marker of coal smoke.122 From clinic records, we also collected information on
anemia, pre-existing and gestational hypertension and diabetes, placental disorders and TORCH
67
(Toxoplasmosis, Others [syphilis, varicella-zoster, parvovirus B19], Rubella, Cytomegalovirus,
Herpes) infections during pregnancy.
Data analysis
We used linear and logistic regression to assess associations between cadmium exposure and fetal
growth, after excluding three participants that reported active smoking during late pregnancy. The
distribution of blood cadmium was skewed by four outliers so we log-transformed the
concentrations. We chose log2 transformations to reduce the influence of extreme values on
regression coefficients and to allow easy interpretation of regression coefficients.24 We adjusted
models of birth weight, length, head circumference, ponderal index and LBW for the following
co-variates: maternal age at birth (<25, 25-29, 30-34, ≥35years), monthly household income
(<600,000 Tugriks, 600,000 to <1,200,000 Tugriks, ≥1,200,000 Tugriks), pre-pregnancy BMI
(continuous), anemia status (anemia, no anemia), sex of the baby, gestational age at birth and
gestational age at birth squared (weeks, continuous), living with a smoker (yes, no), coal stove
density (gers/hectare, continuous), and intervention status (control, intervention). Models of SGA
were adjusted for this same list of co-variates, excluding gestational age, gestational age squared,
and sex, while models of PTB were adjusted for all co-variates excluding gestational age and
gestational age-squared. We used stratified models and interaction terms to evaluate effect
modification by sex of the baby, living with a smoker, coal stove density, and intervention status.
Results are presented per doubling of cadmium exposure for crude and adjusted models for all
births and term births (≥37 weeks).
To investigate coal smoke as a source of maternal cadmium exposure, we regressed log2-
transformed blood cadmium concentrations on coal stove density, while adjusting for the following
variables: age of mother (<25, 25-29, 30-34, ≥35 years), monthly household income (<600,000
Tugriks, 600,000 to <1,200,000 Tugriks, ≥1,200,000 Tugriks), pre-pregnancy BMI (continuous),
anemia status (anemia, no anemia), living with a smoker (yes, no), season in which the blood
sample was collected (winter, spring, summer, fall), and intervention status (control, intervention).
68
4.4. Results
Blood samples were collected from 378 of the 465 participants who had a live birth in the full
UGAAR cohort. One participant who had a child with a chromosomal abnormality and an
additional three participants who reported active smoking in late pregnancy were excluded, leaving
374 participants in our final dataset. Participants with and without a blood cadmium measurement
had similar demographics and lifestyles, including age at enrollment, household income, maternal
education, marital status, and pre-pregnancy BMI. Birth weight, length, and head circumference
were also similar between these groups (Table D.1). However, participants with a cadmium
measurement were recruited earlier in pregnancy (10.0 vs. 12.0 weeks; p=0.003), gave birth later
in pregnancy (39.5 vs. 39.3 weeks; p=0.01) and were less likely to have a baby that was LBW
weight (4% vs 9%; p=0.04) or preterm (5% vs 17%; p=0.001) compared with those who did not
provide a blood sample.
Participants with a blood cadmium measurement were enrolled at a median (25th, 75th percentile)
age of 29.0 years (25.0, 33.0 years) and a gestational age of 10.0 weeks (8.0, 12.0 weeks). Most
participants (86%) were married or in common-law relationships, completed university (81%), and
44% reported a monthly household income in the highest income bracket of ≥1,200,000 Tugriks
(approximately $487 USD) (Table 4.1). Close to half (43%) of participants reported living with a
smoker at any time in pregnancy. Seventy-six percent of participants reported taking iron, folate,
calcium, or multivitamin supplements during pregnancy, but data for 24% of participants were
missing. Few participants had anemia (19%), diabetes (0%), gestational diabetes (0%), gestational
hypertension (8%), or TORCH infections (2%). The median (25th, 75th percentile) birth weight
and gestational age were 3500 (3165, 3800) g and 39.5 (38.5, 40.0) weeks, respectively (Table 1).
The median (25th, 75th percentile) birth length, head circumference and ponderal index were 51 cm
(50, 52 cm), 35 cm (34, 36 cm), 2.6 g/cm3 (2.5, 2.8 g/cm3) , respectively.
The median (25th,75th percentile) blood cadmium concentration was 0.20 µg/L (0.15, 0.29 µg/L).
In bivariate analyses, cadmium concentrations were marginally higher among women who
69
reported living with a smoker (p=0.08) and significantly higher among those living in an apartment
surrounded by a higher density of coal stoves (p=0.02; Table 4.1). Blood cadmium concentrations
were also highest among participants who delivered babies in the lowest birth weight tertile
(p=0.02). Participants who had a blood sample collected in spring (p=0.03) and those in the
intervention group (p=0.008) had lower blood cadmium concentrations.
70
Table 4.1 Maternal blood cadmium concentrations (µg/L) by maternal and newborn
characteristics
N (%)
Median
(25th, 75th percentile) p-value
MATERNAL Maternal age < 25 years 79 (21) 0.20 (0.15, 0.30) 0.49
25-29 years 118 (32) 0.20 (0.15, 0.27) 30-34 years 129 (34) 0.19 (0.14, 0.29) >34 years 49 (13) 0.22 (0.18, 0.31) Maternal education Completed university 302 (81) 0.20 (0.15, 0.29) 0.17
Did not complete university 42 (11) 0.21 (0.16, 0.37) Marital status Married/common-law 320 (86) 0.20 (0.15, 0.29) 0.25
Not married/common-law 54 (14) 0.20 (0.16, 0.31) Monthly household income < 600,000 Tugriksa 92 (25) 0.16 (0.11, 0.29) 0.29
600,000 to <1,200,000 Tugriks 105 (28) 0.22 (0.16, 0.32) ≥ 1,200,000 Tugriks 163 (44) 0.19 (0.15, 0.28) Parity 0 39 (10) 0.22 (0.15, 0.31) 0.68
1 141 (38) 0.19 (0.15, 0.27) ≥2 82 (22) 0.19 (0.12, 0.32) Missing 112 (30) 0.20 (0.16, 0.29) Pre-pregnancy BMI < 20 104 (28) 0.20 (0.15, 0.27) 0.87
20-22 134 (36) 0.20 (0.15, 0.31) ≥ 23 110 (29) 0.19 (0.14, 0.29) Worked outside the home during
pregnancy No 81 (22) 0.22 (0.15, 0.34) 0.20
Yes 281 (75) 0.19 (0.15, 0.28) Anemia status No 304 (81) 0.20 (0.15, 0.29) 0.71
Yes 70 (19) 0.19 (0.15, 0.27) Took iron, folate, calcium, or multi-
vitamin supplements during pregnancy No 0 (0) NA 0.80
Yes 283 (76) 0.19 (0.15, 0.29) Missing 91 (24) 0.20 (0.15, 0.31) Lived with a smoker in late pregnancy No 193 (52) 0.19 (0.15, 0.27) 0.08
Yes 160 (43) 0.21 (0.15, 0.33)
71
Season in which blood sample was
collected Winter (December, January, February) 71 (19) 0.20 (0.16, 0.29) 0.03
Spring (March, April, May) 119 (32) 0.17 (0.14, 0.26) Summer (June, July, August) 102 (27) 0.21 (0.16, 0.30) Fall (September, October, November) 82 (22) 0.23 (0.15, 0.30) Coal stove density (within 5000 m buffer
of apartment) < 3.5 gers/hectare 127 (34) 0.18 (0.14, 0.25) 0.02
3.5-4.5 gers/hectare 130 (35) 0.20 (0.16, 0.3) >4.5 gers/hectare 115 (31) 0.21 (0.16, 0.31) Intervention status Control 173 (46) 0.22 (0.16, 0.31) 0.008
Intervention 201 (54) 0.19 (0.14, 0.27)
NEWBORN Sex of baby
Girls 169 (45) 0.20 (0.15, 0.29) 0.64
Boys 205 (55) 0.19 (0.15, 0.29)
Birth weight
< 3400 g 133 (35) 0.22 (0.16, 0.32) 0.02
3400-3700 g 123 (33) 0.19 (0.15, 0.27)
> 3700 g 118 (32) 0.19 (0.14, 0.27)
Low birth weight No 359 (96) 0.19 (0.15, 0.29) 0.85
Yes 15 (4) 0.22 (0.14, 0.38)
Small for gestational age No 347 (93) 0.19 (0.15, 0.29) 0.16
Yes 27 (7) 0.25 (0.16, 0.32)
Preterm birth No 355 (95) 0.20 (0.15, 0.29) 0.52
Yes 19 (5) 0.20 (0.14, 0.25) aAt the time of data collection, 600,000 Tugriks was equivalent to approximately $243 USD.
Blood cadmium concentrations were associated with decreased birth weight in both crude and
adjusted models (Table 4.2). In adjusted models, a doubling of blood cadmium concentration was
associated with an 86 g (95% CI: 26, 145 g) and 84 g (95% CI: 27, 142 g) decrease in birth weight
among all births and term births, respectively. A doubling in blood cadmium was also associated
with a decrease in ponderal index (-0.04 g/cm3; 95% CI: -0.08, 0.01 g/cm3) among term births and
72
there was a trend toward increased risk of small for gestational age (OR=1.52; 95% CI: 0.93, 2.49).
No other associations with fetal growth or duration of gestation were found.
Table 4.2 Effect of a doubling of maternal blood cadmium concentrations on fetal
growth outcomes
Crude Adjusteda
Outcome
Type of
effect
estimateb
All births
(n = 374)
Term birthsc
(n =355)
All births
(n = 324)
Term birthsc
(n = 311)
Birth weight, g
Mean
difference
(95% CI)
-71
(-136, -7)
-86
(-143, -28)
-86
(-145, -26)
-84
(-142, -27)
Birth length, cm -0.21
(-0.52, 0.10)
-0.21
(-0.46, 0.05)
-0.17
(-0.45, 0.11)
-0.15
(-0.42, 0.12)
Head
circumference,
cm
-0.03
(-0.25, 0.19)
-0.09
(-0.28, 0.10)
-0.09
(-0.29, 0.10)
-0.09
(-0.29, 0.10)
Ponderal index,
g/cm3
-0.01
(-0.05, 0.02)
-0.03
(-0.06, -0.001)
-0.04
(-0.08, 0.002)
-0.04
(-0.08, -0.01)
Low birthweight
Odds ratio
(95% CI)
0.90
(0.47, 1.72)
1.85
(0.72, 4.79)
1.16
(0.48, 2.80)
1.57
(0.60, 4.13)
Small for
gestational age
1.43
(0.90, 2.26)
1.45
(0.87, 2.42)
1.52
(0.93, 2.49)
1.45
(0.84, 2.50)
Preterm birth 0.66
(0.36, 1.21)
1.52
(0.80, 2.87)
-----
aModels of birth weight, birth length, head circumference, ponderal index and low birth weight were adjusted for
maternal age, monthly household income, pre-pregnancy BMI, anemia status, sex of the baby, gestational age and
gestational age squared, living with a smoker in late pregnancy, ger density, and intervention status. Models of small
for gestational age were adjusted for the same list of variables, excluding gestational age, gestational age squared, and
sex, and models of preterm birth were adjusted for the same list of variables, excluding gestational age and gestational
age squared. c≥37 weeks gestation
After stratifying by sex of the baby, a significant effect on birth weight was only seen for girls
(Figure 4.1), although an interaction between cadmium and sex was not significant (p=0.45).
Similarly, living with a smoker, coal stove density, and intervention status were also not significant
effect modifiers of the relationship between cadmium exposure and birth weight (Figure 4.1) or
other outcomes (Tables D2-D5).
73
Figure 4.1 Estimated effects of a doubling of maternal blood cadmium (Cd)
concentrations on birth weight in stratified analyses
After adjustment in a multiple linear regression model, an interquartile increase in coal stove
density (from 3 to 5 gers/hectares) was associated with an 11% (95% CI: 0, 23%) increase in blood
cadmium concentrations. Living with a smoker was associated with a 12% (95% CI: 0, 25%)
increase in blood cadmium.
4.5. Discussion
This observational study was nested within the UGAAR randomized controlled trial of portable
HEPA cleaners and early-life growth and development in Ulaanbaatar, Mongolia. We found that
74
cadmium exposure during pregnancy was associated with reductions in birth weight in adjusted
models. We previously reported higher blood cadmium concentrations among participants who
lived with a smoker as well as 14% (95% CI: 4, 23%) lower blood cadmium concentrations among
intervention participants using HEPA cleaners during pregnancy, indicating an airborne source of
exposure.55 Here we found that the density of coal stoves surrounding participants’ apartments was
associated with blood cadmium concentrations.
Birth weight was reduced by 86 g (95% CI: 26, 145 g) per doubling of maternal blood cadmium
concentrations in adjusted models, an exposure contrast that roughly corresponds with the blood
cadmium interquartile range of 0.15-0.29 µg/L. Others have also reported decreases in birth weight
with gestational cadmium exposure.11-13,15,17,20,60,63,69 Taylor et al. (2016) reported a 63 g (95% CI:
18, 107 g) decrease in term birth weight per 1 µg/L increase in maternal blood cadmium
concentrations among 4,191 pregnant women living in the United Kingdom.13 Vidal et al. (2015)
reported a 52 g decrease in birth weight with each log-unit increase in maternal blood cadmium
concentrations (se=24.2, p=0.03) among 276 pregnant women in North Carolina, US.19 Maternal
urinary cadmium concentrations have also been linked to significant decreases in birth
weight.11,20,25,72 However, the evidence is not consistent, with others reporting no associations
between maternal blood or urinary cadmium concentrations and birth weight.16,24,67,71,173
After stratifying by sex of the baby, maternal blood cadmium concentrations were associated with
significant decreases in birth weight for girls only, but sex was not a significant effect modifier.
Others have reported greater effects of cadmium exposure among girls,11,13,17,20,22,25 and the lack
of stratified analyses in some studies has been suggested as a potential explanation for the mixed
findings of cadmium effects on fetal growth.13 Taylor et al. (2016) reported significantly larger
effects among girls than boys for birth weight, head circumference, and birth length.174 Similarly,
Cheng et al. (2016) reported that each log unit increase in maternal urinary cadmium (µg/g
creatinine) was associated with a 117g (95% CI: 25, 209g) decrease in birth weight among girls,
with no significant effects among boys.11 The lack of a significant interaction effect seen in our
study may be due, in part, to insufficient power to evaluate interactions.
75
There are multiple plausible pathways through which cadmium may affect fetal growth. Cadmium
is absorbed into the body via inhalation or ingestion; roughly 10-50% of inhaled cadmium is
absorbed compared with only 3-5% through ingested cadmium.51,175 Cadmium is primarily stored
in the kidneys and liver,51 but during pregnancy it can also accumulate in the placenta and cross
into the fetal environment.82 Once in the placenta, cadmium can interfere with the transfer of zinc
and other essential nutrients,59,82 as well as with placental release of progesterone.83 Cadmium may
also induce oxidative stress, which can cause direct cellular damage and initiate secondary
processes, including systemic inflammation, endothelial dysfunction, and increased blood
pressure, ultimately decreasing utero-placental blood flow.170,176 Cadmium has been linked to an
increased risk of gestational hypertension in pregnant women, a risk factor for poor fetal
growth.29,177 Cadmium has also been shown to inhibit the release of the placental enzyme 1β-
hydroxysteroid dehydrogenase type 2 (1β-HSD2), which is responsible for protecting the fetus
from excess maternal cortisol.77,178 Low enzymatic activity of 1β-HSD2 has been associated with
increased risks of fetal growth restriction and preterm birth.77 The sex-selective effects of cadmium
may be, in part, due to differential responses in the methylation of growth-related genes, with
hypomethylation in girls and hypermethylation in boys.13
Much of the research investigating air pollution effects on fetal growth has focussed on PM2.5 as
the main pollutant of interest. Studies have found relatively small but consistent effects of PM2.5
on birth weight, LBW, SGA, and PTB.4,7 In a recent meta-analysis of 17 studies, authors reported
that each 10 µg/m3 increase in outdoor PM2.5 was associated with a 16 g (95% CI: 5, 27 g) decrease
in birth weight.4 Both the size and composition of PM2.5 particles influence their toxicity, but few
studies have attempted to understand how differences in composition may affect fetal
growth.4,179,180 In an investigation of PM2.5 and LBW in 22 US counties, Hao et al. (2015)
suggested that differences in nitrate and sulphate concentrations may in part explain differences in
county-level effect estimates.180 Similarly, Sun et al. (2016) reported that birth weight has been
negatively associated with zinc, nickel, titanium, and vanadium particles, in their meta-analysis of
outdoor PM2.5 and fetal growth; none of the studies included in the analysis investigated cadmium.4
76
Differences in oxidative potential has been suggested to modify the relationship between PM2.5
and fetal growth,181 which in turn is attributed to its compositional differences in metals, polycyclic
aromatic hydrocarbons (PAHs), and other components.182,183 PAHs have been identified as
particularly toxic components of coal-, tobacco-, and traffic-related PM2.5 emissions that are linked
to impaired fetal growth.184,185 Cadmium may be another component of PM2.5 mixtures that can
explain, in part, the detrimental effects of PM2.5 on fetal growth in some settings. Like cadmium,
PM2.5 has also been shown to induce oxidative stress, inflammation, and endothelial dysfunction155
and these overlapping pathways may contribute to increased particle toxicity on fetal growth.
We interpret our finding of an association between coal stove density and blood cadmium as
evidence that coal smoke is a source of cadmium among this population. We previously reported
that the locations of ger neighbourhoods in Ulaanbaatar explained 66% of the variability in outdoor
SO2 concentrations, and that coal stove density was predictive of indoor PM2.5 concentrations in
the UGAAR cohort.122 Notably, the impact of an interquartile range increase in coal stove density
on blood cadmium (11%, 95% CI: 0, 23%) was similar to that of living with a smoker during
pregnancy (12%, 95% CI: 0, 25%). SHS is an established source of cadmium.17,186,187
Source apportionment studies have linked industrial coal use to outdoor air cadmium
concentrations in parts of China, New Zealand, and the US.171,188-191 However, few studies have
investigated industrial or residential coal emissions as a contributor to blood or urinary cadmium
concentrations in communities that rely heavily on coal, particularly among pregnant women. Two
studies reported that blood cadmium concentrations did not significantly differ in pregnant women
who lived near a coal combustion factory or used coal as cooking fuel compared with those that
were not exposed to coal, in China (n=215) and South Africa (n=641).17,74 In contrast, Zhang et al.
(2016) reported significantly higher airborne cadmium concentrations in homes using coal for
cooking and heating (n=12) versus gas or electricity (n=83), as well as significantly higher
cadmium concentrations during the heating (February to March) vs non-heating (April to January)
season in Lanzhou, China.75 Authors also reported a 3 g decrease in birth weight per 1 ng/m3
increase in residential airborne cadmium concentration in the heating season only (p=0.05).75 The
77
mixed study findings may in part be explained by differences in coal composition, which can
depend on the type of coal as well as the geographic area from which it is mined.192-194
We cannot rule out the possibility that the association between coal stove density and cadmium
exposure is confounded by diet. Diet is considered the most important source of cadmium among
non-smokers,50,51,53,195 but its relative importance to total exposures will depend on the population
and setting. In general, diet-related exposures may be higher in areas where cadmium levels are
naturally high, or where industrial or agricultural activities have contaminated surrounding soil.
For example, diets high in fruits and vegetables are estimated to be a major contributor to cadmium
exposure in Bangladesh due to widespread fertilizer-related soil contamination,196 while diet has
been found to be a relatively modest contributor to total exposures compared with active smoking
among other populations, including among Norwegian women aged 28-40 years and the general
Canadian population.53,197 Limited information exists on the consumption of cadmium-rich foods
in Ulaanbaatar, such as leafy green vegetables, shellfish, and organ meats. Some evidence suggests
that women who migrated to Ulaanbaatar from the countryside may be more likely to eat a
traditional diet consisting primarily of dairy and meat products.198 Our questionnaires included
questions about migration to Ulaanbaatar, but response rates for these questions were poor.
Families who migrate from the country-side are more likely to be of lower income since most
move in search of economic opportunities, and therefore, may be more likely to live in less
expensive ger neighborhoods than apartments.100 Ours was a relatively wealthy study population,
and our models were adjusted for self-reported household income.
Blood cadmium concentrations measured in our study population were relatively low, with a
median (25th, 75th percentile) concentration of 0.20 µg/L (0.15, 0.29 µg/L). In contrast, Wang et
al. (2016) reported a median (25th, 75th percentile) blood cadmium of 0.80 µg/L (0.57, 1.06 µg/L)
in 3,254 non-smoking pregnant women living in six Chinese cities. No specific source of cadmium
was investigated in that study.14 Similarly, in a review of 24 studies assessing blood cadmium
concentrations among smoking and non-smoking pregnant women, Taylor et al. (2014) reported
mean and median blood cadmium concentrations ranging from 0.09 to 2.26 µg/L among
78
populations residing in Poland, Russia, South Africa, Egypt, India, Norway, France, United States
and China.138
Some limitations of our study should be noted. Our study sample was relatively small, which
limited our ability to evaluate effect modifiers. The full UGAAR cohort consisted of 465 live births
of which 374 non-smoking participants with an available cadmium blood measurement were
included in this analysis. Diet may have potentially confounded the relationship seen between coal
stove density and birth weight, but we were unable to assess consumption of typically cadmium-
rich foods as a source of exposure in our study. However, we reasoned that airborne exposures
were important, as evidenced by the reduction in blood cadmium concentrations among
intervention participants who used HEPA cleaners. We did not assess micronutrient status,
including levels of zinc and selenium, which can affect cadmium uptake and accumulation, while
questionnaire data related to iron, folic acid, calcium, and multivitamin supplementation had poor
response rates. Finally, we assessed blood cadmium concentrations only at one time in pregnancy,
but previous studies have demonstrated that exposures do not change appreciably throughout
pregnancy.199
4.6. Conclusions
Gestational cadmium exposure was associated with decreased birth weight, and neighborhood coal
stove density, a proxy of residential coal smoke, was associated with higher maternal blood
cadmium concentrations, in this cohort of non-smoking pregnant women. In some settings
cadmium may play a role in the putative relationship between air pollution and impaired fetal
growth.
79
Chapter 5.
Discussion
5.1. Summary
In our randomized controlled trial, we evaluated the impact of portable HEPA cleaner use during
pregnancy on residential indoor PM2.5 and maternal blood cadmium concentrations, as well as the
effect of HEPA cleaner use during pregnancy and maternal cadmium exposure on fetal growth.
The main findings from each chapter are summarized below.
Effect of portable HEPA cleaners on indoor PM2.5 concentrations and SHS
exposure (Chapter 2)
HEPA cleaner use was associated with a 29% (95% CI: 21, 37%) lower mean indoor PM2.5
concentration. Effectiveness was highest in winter when the geometric mean indoor PM2.5
concentrations were reduced from 45 to 29 µg/m3. Effectiveness was greater when the HEPA
cleaners were first deployed (40%, 95 % CI: 31, 48%) than after roughly five months of use (15%,
95 % CI: 0, 27%). We also observed greater reductions in PM2.5 concentrations in homes with two
HEPA cleaners (33%, 95% CI: 25, 41%) versus one cleaner (20%, 95% CI: 6, 32%). The density
of air cleaners per area of the apartment was similar between the homes, so the difference in
effectiveness may have been, in part, due to differences in use. For example, we heard anecdotal
reports that some participants turned air cleaners off at night due to noise. For homes with two
HEPA cleaners, participants may have turned off the unit in the main bedroom and allowed the
second unit, which was located in the main living area, to run continuously. We found no evidence
of effect modification by air cleaner density, self-reported air cleaner use, season, or window
opening. Blood cadmium concentrations were, on average, 14% (95% CI: 4, 23%) lower among
intervention participants. We reasoned that SHS was a source of airborne cadmium among our
non-smoking population based on the following: (1) close to half of all study participants lived
with smokers indicating SHS exposure was common, and (2) we found higher blood cadmium and
hair nicotine concentrations as well as stronger correlations between the two measures among
80
participants who lived with a smoker. Overall, we concluded that HEPA cleaners are an effective
intervention to reduce residential indoor PM2.5 concentrations and SHS exposure among pregnant
women living in a highly polluted community.
Effect of HEPA cleaner use on fetal growth (Chapter 3)
We found that HEPA cleaner use was not associated with fetal growth outcomes in our main
analysis, but in a pre-specified subgroup analysis that was limited to 429 term births, we found
that their use was associated with greater birth weight of 85 g (95% CI: 3, 167 g). Surprisingly, we
found a lower risk of spontaneous abortions (OR=0.38; 95% CI: 0.18, 0.82) among the intervention
group. In contrast, we also found a higher frequency of PTB (10% vs 4%; p=0.03) in the
intervention group, and we reasoned that this could have offset the beneficial effect of the
intervention among all births. We found a 100 g difference in median birth weight between the
control (3450 g) and intervention (3550 g) participants, and when we adjusted for PTB we found
that the estimated intervention effect on birth weight (84 g; 95% CI: -1, 170 g) was similar to that
from the subgroup analysis on term births. We speculated that the higher risk of PTB and the lower
risk of spontaneous abortion may have been explained by the intervention allowing for more
fetuses to survive long enough to be born preterm. We found no evidence of effect modification
by living with a smoker, average time spent indoors at home during pregnancy, sex of the baby,
season of birth, self-reported air cleaner use, or household income. Overall, we concluded that use
of HEPA cleaners during pregnancy was associated with greater term birth weight.
Effect of gestational cadmium exposure on fetal growth (Chapter 4)
We found that a doubling of maternal blood cadmium concentrations was associated with an 86 g
(95% CI: 26, 145 g) and 84 g (95% CI: 2, 142 g) decrease in birth weight and term birth weight,
respectively, in models adjusted for maternal age, monthly household income, pre-pregnancy BMI,
anemia status, sex of the baby, gestational age and gestational age at birth squared, living with a
smoker, coal stove density within 5000 m radius of the participant’s apartment, and intervention
81
status. A doubling of blood cadmium concentrations was also associated with decreased ponderal
index (-0.04 g/cm3; 95% CI: -0.08, -0.01 g/cm3) among term births only, in adjusted models. We
found no evidence of effect modification by living with a smoker, coal stove density, intervention
status, or sex of the baby. We also found that coal stove density, a proxy for coal smoke,
contributed to maternal blood cadmium concentrations. An interquartile increase in coal stove
density (from three to five gers/hectare) was associated with an 11% (95% CI: 0, 23%) increase in
blood cadmium concentrations, after adjusting for age of mother, monthly household income, pre-
pregnancy BMI, anemia status, living with a smoker, season in which blood sample was collected,
and intervention status. Overall, we concluded that cadmium exposure during pregnancy is
associated with decreased birth weight, and that in some settings, cadmium may play a role in the
putative relationship between air pollution and impaired fetal growth.
5.2. Synthesis and significance of findings
We used HEPA cleaners as a tool to study the effects of residential indoor PM2.5 on fetal growth.
We showed that lowering indoor PM2.5 concentrations during pregnancy led to higher term birth
weight by an average of 85 g (95% CI: 3, 167 g). An 85 g effect may not be clinically relevant,
but it can be meaningful on a population level since a rightward shift in the birth weight distribution
of a population will result in fewer babies born at lower birth weights. Birth weight is an indicator
of impaired fetal growth, a condition that can affect multiple systems in the body. Moreover,
unfavorable intrauterine conditions are thought to force the fetus to make irreversible adaptions to
ensure immediate survival. These adaptions are thought to have lasting effects on organ
morphology, vasculature, physiology, and endocrine and metabolic functioning.39,200 Evidence
suggests that persons born growth restricted are at increased risks of poorer neurodevelopmental
outcomes and obesity in childhood and of type-2 diabetes, hypertension and coronary heart disease
in adulthood.26,31-35
Our finding of improvements in fetal growth with reductions in air pollution is consistent with the
bulk of observational studies, which have found links between PM2.5 concentrations during
82
pregnancy and decreased birth weight and other indicators of impaired fetal growth.4-10,43-46 A
recent meta-analysis of observational studies reported a 16 g (95% CI: 5, 27 g) reduction in birth
weight and an OR of 1.09 (95% CI: 1.03, 1.15) for LBW per 10 µg/m3 increase in PM2.5 over full
pregnancy.4 The majority of studies included in the paper limited their analyses to term births to
evaluate the effect of air pollution on “normal” fetal growth.4,28
A limited number of studies show detrimental effects of PM and other pollutants on fetal growth
in LMICs.47,201-204 Sources and PM2.5 composition may differ between higher and lower income
countries, which may in turn influence particle toxicity. Coal-related PM2.5, which is more
common in LMICs,205 has been associated with increased risks of ischemic heart disease and
mortality, with some evidence suggesting that coal smoke particles may be more toxic than those
from other sources.206-208 Our study was conducted in a community that relies heavily on
residential coal use that contributes to 45-70% of total PM2.5 concentrations in the city.99,100 In
addition to PM2.5, coal smoke contains other pollutants that have been linked to impaired fetal
growth, including cadmium, lead, mercury, and PAHs.58,209-212 We showed that the density of coal
stoves surrounding the apartments of pregnant women contributed to cadmium exposure and that
cadmium was linked to decreased birth weight. Coal smoke has been linked to impaired fetal
growth by others. A 2014 meta-analysis of 19 studies estimated that unprocessed solid fuel
combustion in homes, including coal, reduced mean birth weight by 86 g (95% CI: 56, 117 g) and
increased risks of LBW (OR=1.35; 95% CI: 1.23, 1.48).213 Another study reported a 1.1% (95%
CI: 0.2, 2.0%) increase in LBW for each 5-km closer that maternal residences were to coal and
solid waste power plants in Florida, US.214 PM2.5 and cadmium share mechanistic pathways,
including oxidative stress and inflammation, which could potentially increase the toxicity of coal-
related PM2.5 emissions.
Interventions are needed to address the air pollution-related burden of disease. In 2016, 95% of
the world’s population lived in areas where outdoor PM2.5 concentrations exceeded the World
Health Organization’s annual average guideline of 10 µg/m3, and concentrations are increasing in
much of the world, particularly in LMICs.2 Consequently, the burden of disease is likely to
83
continue to increase if actions are not taken to reduce exposures at community and household
levels.
Previous research shows that HEPA cleaners can reduce residential indoor PM2.5 concentrations
by 25 to 79%, in relatively low pollution settings and over relatively short periods of time.87,89-95
Our work adds to this literature by showing that HEPA cleaners can also lower residential indoor
PM2.5 concentrations in highly polluted communities over a period of several months. Other factors
also make HEPA cleaners an effective intervention. HEPA cleaners are relatively inexpensive and
easy to use, allowing individuals some control over reducing their air pollution exposures. HEPA
cleaners address exposures in indoor settings where individuals spend a majority of their time,
with exposure reduction benefits being available to all occupants in these settings. HEPA cleaners
also lower concentrations of indoor-generated pollutants and those that infiltrate from outdoors.
This intervention can be useful in multiple settings, including in LMICs where populations are
overburdened by increasing air pollution concentrations and other environmental stressors,215 in
communities located near busy roadways, industry, or other sources,95 as well as those
experiencing temporary poor air quality events. The use of portable air cleaners is currently
recommended to communities affected by wildfire smoke by several public health agencies,
including the US Environmental Protection Agency and the BC Centre for Disease Control.216-219
However, HEPA cleaners are only a near-term solution to addressing air pollution exposures, and
even then, are limited in the benefits they can provide. For example, HEPA cleaners operated in
the home will have no impact on exposures experienced in other microenvironments, including
during a commute, at work, or outdoors. Their effectiveness over the long term has not been well
studied, with our study being the first to investigate longer term air cleaner use in a highly polluted
setting. One of the biggest drawbacks of recommending HEPA cleaner use in homes is that it
places the burden of exposure reduction on the individual or household even though the pollution
may be mainly due to community sources over which individuals have little control. Lower income
persons may not be able to access or may not benefit from HEPA cleaners for several reasons.
HEPA cleaners have initial purchase, operating, and maintenance costs that may be prohibitive to
84
some households. Air cleaners also need a constant and reliable supply of electricity and should
be used in homes that are relatively well constructed, since air cleaners are more effective at lower
air exchange rates. In Ulaanbaatar, residents of ger households are the most highly exposed
group,100 but HEPA cleaners would probably have limited value in these homes due to a lack of
electricity and higher air exchange rates in gers. Additionally, air cleaners may be less effective in
warm climates if residents are unable to limit air exchange rates by keeping windows and doors
closed.
Our findings suggest that HEPA cleaner use during pregnancy can benefit fetal growth. However,
public health messaging on their use needs to be carefully crafted. Pregnant women and their
fetuses are especially vulnerable to air pollution, but birth does not mark the end of the vulnerable
period. Newborns, infants, and young children are also vulnerable,220,221 and recommendations
need to convey that extending the use of HEPA cleaners beyond pregnancy is likely most
beneficial. Households should also be made aware that air cleaners are most effective when air
exchange rates are limited in the home, as well as about potential risks from heat and indoor-
generated pollutants when homes are kept tightly sealed. Finally, information also needs to be
provided on appropriate sizing of HEPA cleaners in the rooms in which they are to be used, as
well as on maintenance and timely replacement of filters.
The only way to effectively address air pollution exposures over the long term at the population
level is to implement interventions that reduce air pollution emissions. Stricter emissions
regulations, improvements in technology, and a shift toward cleaner fuels in high-income countries
in recent decades have allowed for cost-effective decreases in air pollution and large public health
benefits. In the US, programs implemented under the 1990 Clean Air Act are projected to lead to
substantial long-term air quality improvements and reductions in premature death and illness.222
The economic value of these improvements is estimated to reach $2 trillion by 2020, which far
exceeds the costs of the program, which are estimated at $65 billion by 2020.222 Reductions in
PM2.5 have also been associated with decreased mortality in other parts of the world, including
Europe, Asia, and Australia.223-226
85
Multiple strategies have been proposed to curb increasing air pollution concentrations in
Ulaanbaatar, including a shift to cleaner-burning fuels in ger households, improved emission
controls for coal-fired power plants, and cleaner busses and cars.97 Programs to disseminate
cleaner-burning coal stoves in ger households have had modest success in reducing outdoor PM2.5
concentrations.227 Without aggressive actions, air pollution concentrations will likely increase as
the city sees growth in the number of ger households and motor vehicles, and as the demands for
energy from power plants increase.100,101 Implementing long-term solutions to reduce air pollution
emissions in any community will take several decades,109 but such actions will lead to long-term
exposure reduction benefits for all residents.
5.3. Strengths and limitations
The key strength of our study is the randomized study design. The bulk of the current research
consists of observational studies, that while valuable, are limited in how much they can tell us
about the relationship between air pollution and fetal growth, while randomized trials are scare.
We minimized confounding by randomizing participants to control and intervention groups. We
used the intervention as a tool to study the link between air pollution and fetal growth by
introducing a concentration gradient in PM2.5 concentrations between the groups. We assessed
indoor PM2.5 concentrations, collecting 447 measurements in homes, in contrast to observational
studies which have largely focused on outdoor air pollution measurements. Our assessment was
the first to evaluate the impact of HEPA cleaner use over a relatively long duration in a highly
polluted community. As mentioned earlier, this is the second randomized trial of air pollution and
fetal growth; our sample size of 463 live births was larger than the previous study which included
174 births.144 Finally, we assessed coal smoke as a potential cadmium source among pregnant
women, which has not been adequately investigated in the literature.
86
There are also several limitations of our study. First, participants were not blinded to their
intervention status. Although our outcome measures were objective and therefore, unlikely to have
been biased, the lack of blinding likely contributed to the greater loss to follow up among the
control group (19 versus nine participants). Participants received air cleaners at a median of 11
weeks gestation so the intervention did not influence exposure during much of the first trimester.
Our estimates of HEPA cleaner impacts on PM2.5 concentrations were based on two one-week
monitoring periods in early and later pregnancy, and it is possible that participants used the air
cleaners more frequently during these periods, which would have caused us to overestimate the
concentration reductions introduced by the intervention. Due to mechanical issues, our assessment
of air cleaner effectiveness was limited because we were unable to objectively quantify air cleaner
use; instead, we had to rely on self-reported use. However, neither measure allows for an
assessment of changes in use since we were unable to assess whether HEPA cleaner use was
different in earlier versus later stages of the study period, or whether use differed during day- and
night-time periods. We did not systematically assess reasons that participants failed to use the
intervention continuously throughout the study period, including issues around noise. There was a
high frequency of missing data collected via questionnaires, including questions on parity, the
number of smokers in the home, previous birth complications, and use of supplements. Finally,
time activity data was crudely assessed; participants were asked to provide a daily breakdown for
time spent in different microenvironments over a “typical” week in early and late pregnancy.
5.4. Future research directions
Several gaps remain in our understanding of the relationship between air pollution exposure and
fetal growth that merit further investigation. The potential joint effects of multiple pollutants on
fetal growth have been poorly studied, while we know little about the shape of the exposure-
response relationship between PM2.5 and fetal growth outcomes, whether a “threshold” or “safe”
exposure level exists, how composition of PM2.5 affects toxicity, and critical exposure windows.
There is also a lack of studies investigating interventions aimed at reducing air pollution exposures
during pregnancy.
87
Much of the available literature has relied on single pollutant models that do not consider potential
joint effects of multiple pollutants. Pregnant women are exposed to multiple environmental
contaminants228 that together may impair fetal growth but few studies have attempted to investigate
how multiple pollutants may influence fetal growth.229,230 Similarly, few studies have investigated
the effects of specific sources or particle composition on fetal growth.4,180,214 We showed that both
PM2.5 and cadmium can impair fetal growth and reasoned that in some settings, cadmium may play
a role in the putative relationship between air pollution and impaired fetal growth, in part due to
overlapping mechanistic pathways. Others have linked living near industrial sources or busy
roadways, as well as exposure to specific contaminants in PM2.5 mixtures, such as zinc, vanadium,
nitrates, and copper, to increased toxicity.4,179,180,214,231
Another important question that remains is the shape of the exposure-response relationship
between PM2.5 and fetal growth. Although non-linear functions have been identified for other
PM2.5-related health outcomes, including cardiovascular mortality and some respiratory effects,232-
234 analyses of PM2.5 and fetal growth have mostly assumed linear functions. One of the main
challenges in assessing the shape of the function is that most studies have been conducted in high-
income countries where concentrations are relatively low. Understanding the shape of the
exposure-response function is necessary to assess population-level impacts of air pollution on fetal
growth as well as to understand the potential benefits of exposure reduction at different
concentrations. A related issue is whether an effect threshold exists. PM2.5-related effects on
respiratory and cardiovascular morbidity and mortality appear to have no thresholds,221 and it may
be reasonable to assume that no threshold exists for fetal growth outcomes considering that effects
have been reported even in settings where mean outdoor PM2.5 concentrations were between 5-
10 µg/m3.235
Critical exposures windows are also not well understood. Some evidence suggests that exposures
in the latter half of pregnancy are most detrimental.4 Evidence linking air pollution exposures with
repeated ultrasound measurements suggests that exposures may not alter growth trajectories until
88
late in the second trimester.153,154 Another study reported that babies whose eighth month of
gestation overlapped during the 2008 Beijing Olympics, when air pollution concentrations were
substantially reduced, had a 23 g (95% CI: 5 g, 40 g) greater mean birth weight compared with
babies born during the same calendar dates the year prior to and after the Olympics.152 We were
unable to investigate exposure windows. In Ulaanbaatar, air pollution concentrations are strongly
associated with season, with the highest concentrations occurring in winter when coal use is
greatest. However, season is also correlated with other factors that impact exposures, including air
exchange in homes and behavior. We found that many participants visited the country-side for
extended periods in the summer, which would have also reduced the impact of the intervention.
Finally, more intervention studies are needed to better understand how reductions in exposure
translate to improvements in health and which interventions are most effective. This type of
information is critical to guide decision making, particularly when interventions are costly and
require commitment from various stakeholders and partners. We showed HEPA cleaners were
effective at reducing indoor PM2.5 and blood cadmium concentrations, but many questions remain
about how to maximize exposure reduction and health benefits, including how factors such as
noise and cost affect uptake and adherence to the intervention and how patterns of use throughout
the day, in different climates, and in different housing types influence effectiveness.
Filling knowledge gaps is important to allow for a fuller understanding of how air pollution affects
fetal growth. However, a lack of a complete understanding of this relationship should not preclude
action since the available research collectively suggests that air pollution exposures are harmful.
Interventions should be implemented to address exposures on community and household levels,
and attention should be given to evaluating these interventions to ensure they are effective in
reducing exposures and health effects, and that benefits are accessible to all community members.
89
5.5. Conclusions
The main aims of this research were to (1) investigate the causal role of PM2.5 on fetal growth, (2)
investigate the relationship between cadmium and fetal growth, and (3) evaluate portable HEPA
cleaners as a household level intervention. Portable HEPA cleaner use during pregnancy is an
effective household intervention to reduce indoor PM2.5 concentrations and SHS exposures and to
improve fetal growth. Our findings reinforce the evidence that reducing air pollution exposure can
benefit health. Many questions still remain on the relationship between air pollution and fetal
growth, including the role of multiple pollutants, how PM2.5 composition affects toxicity, the shape
of the exposure-response function, particularly at higher levels of exposure, and critical windows
of exposure. However, enough evidence exists to warrant immediate actions to lower community-
wide and household level air pollution exposures in the near- and long-term.
90
References
1. Cohen AJ, Brauer M, Burnett R, et al. Estimates and 25-year trends of the global burden
of disease attributable to ambient air pollution: an analysis of data from the Global Burden
of Diseases Study 2015. The Lancet 2017; 389(10082): 1907-18.
2. Health Effects Institute. State of Global Air 2018. 2018.
https://www.stateofglobalair.org/sites/default/files/soga-2018-report.pdf.
3. Landrigan PJ, Fuller R, Acosta NJR, et al. The Lancet Commission on pollution and health.
The Lancet 2017.
4. Sun X, Luo X, Zhao C, et al. The associations between birth weight and exposure to fine
particulate matter (PM) and its chemical constituents during pregnancy: A meta-analysis.
Environmental Pollution 2016; 211: 38-47.
5. Lamichhane DK, Leem JH. A meta-analysis of exposure to particulate matter and adverse
birth outcomes. 2015; 30: e2015011.
6. Sun X, Luo X, Zhao C, et al. The association between fine particulate matter exposure
during pregnancy and preterm birth: a meta-analysis. BMC pregnancy and childbirth 2015;
15: 300.
7. Zhu X, Liu Y, Chen Y, Yao C, Che Z, Cao J. Maternal exposure to fine particulate matter
(PM2.5) and pregnancy outcomes: a meta-analysis. Environmental science and pollution
research international 2015; 22(5): 3383-96.
8. Dadvand P, Parker J, Bell ML, et al. Maternal exposure to particulate air pollution and term
birth weight: a multi-country evaluation of effect and heterogeneity. Environmental Health
Perspectives 2013; 121(3): 267-373.
9. Sapkota A, Chelikowsky A, Nachman K, Cohen A, Ritz B. Exposure to particulate matter
and adverse birth outcomes: a comprehensive review and meta-analysis. Air Quality,
Atmosphere & Health 2012; 5(4): 369-81.
10. Stieb DM, Chen L, Eshoul M, Judek S. Ambient air pollution, birth weight and preterm
birth: a systematic review and meta-analysis. Environ Res 2012; 117: 100-11.
11. Cheng LL. Critical Windows of Prenatal Exposure to Cadmium and Size at Birth.
International journal of environmental research and public health 2016; 14(1): 58.
12. Luo Y, McCullough LE, Tzeng J-Y, et al. Maternal blood cadmium, lead and arsenic
levels, nutrient combinations, and offspring birthweight. BMC public health 2017; 17: 354.
91
13. Taylor CM, Golding J, Emond AM. Moderate Prenatal Cadmium Exposure and Adverse
Birth Outcomes: a Role for Sex-Specific Differences? Paediatric and perinatal
epidemiology 2016; 30(6): 603-11.
14. Wang H, Liu L, Hu Y-F, et al. Maternal serum cadmium level during pregnancy and its
association with small for gestational age infants: a population-based birth cohort study.
Scientific Reports 2016; 6: 22631.
15. Bloom MS, Buck Louis GM, Sundaram R, Maisog JM, Steuerwald AJ, Parsons PJ. Birth
outcomes and background exposures to select elements, the Longitudinal Investigation of
Fertility and the Environment (LIFE). Environmental research 2015; 138: 118-29.
16. Hu X, Zheng T, Cheng Y, et al. Distributions of Heavy Metals in Maternal and Cord Blood
and the Association with Infant Birth Weight in China. The Journal of reproductive
medicine 2015; 60(1-2): 21-9.
17. Röllin HB, Kootbodien T, Channa K, Odland JØ. Prenatal Exposure to Cadmium, Placental
Permeability and Birth Outcomes in Coastal Populations of South Africa. PLoS One 2015;
10(11): e0142455.
18. Thomas S, Arbuckle TE, Fisher M, Fraser WD, Ettinger A, King W. Metals exposure and
risk of small-for-gestational age birth in a Canadian birth cohort: The MIREC study.
Environmental Research 2015; 140(0): 430-9.
19. Vidal AC, Semenova V, Darrah T, et al. Maternal cadmium, iron and zinc levels, DNA
methylation and birth weight. BMC pharmacology & toxicology 2015; 16: 20.
20. Zhang Y, Xu X, Chen A, et al. Maternal urinary cadmium levels during pregnancy
associated with risk of sex-dependent birth outcomes from an e-waste pollution site in
China. Reproductive Toxicology 2018; 75(Supplement C): 49-55.
21. Guo J, Wu C, Qi X, et al. Adverse associations between maternal and neonatal cadmium
exposure and birth outcomes. Science of The Total Environment 2017; 575(Supplement
C): 581-7.
22. Huang K, Li H, Zhang B, et al. Prenatal cadmium exposure and preterm low birth weight
in China. Journal Of Exposure Science And Environmental Epidemiology 2017; 27: 491-
6.
23. Wai MK, Mar O, Kosaka S, Umemura M, Watanabe C. Prenatal Heavy Metal Exposure
and Adverse Birth Outcomes in Myanmar: A Birth-Cohort Study. International Journal of
Environmental Research and Public Health 2017; 14(11).
24. Romano ME, Enquobahrie DA, Simpson C, Checkoway H, Williams MA. Maternal body
burden of cadmium and offspring size at birth. Environmental Research 2016; 147: 461-8.
92
25. Kippler M, Tofail F, Gardner R, et al. Maternal Cadmium Exposure during Pregnancy and
Size at Birth: A Prospective Cohort Study. Environmental Health Perspectives 2012;
120(2): 284-9.
26. Kramer M. The epidemiology of adverse pregnancy outcomes: an overview. J Nutr 2003;
133(5 Suppl 2): 1592S-6S.
27. Mayer C, Joseph KS. Fetal growth: a review of terms, concepts and issues relevant to
obstetrics. Ultrasound in obstetrics & gynecology : the official journal of the International
Society of Ultrasound in Obstetrics and Gynecology 2013; 41(2): 136-45.
28. Wilcox AJ. On the importance—and the unimportance— of birthweight. International
Journal of Epidemiology 2001; 30(6): 1233-41.
29. Adams MM, Alexander GR, Kirby RS. Perinatal Epidemiology for Public Health Practice.
New York: Springer; 2009.
30. Gardosi J. Intrauterine growth restriction: new standards for assessing adverse outcome.
Best Practice & Research Clinical Obstetrics & Gynaecology 2009; 23(6): 741-9.
31. McIntire DD, Bloom SL, Casey BM, Leveno KJ. Birth weight in relation to morbidity and
mortality among newborn infants. The New England journal of medicine 1999; 340(16):
1234-8.
32. Barker DJ. Adult consequences of fetal growth restriction. Clinical obstetrics and
gynecology 2006; 49(2): 270-83.
33. Gluckman PD, Hanson MA, Cooper C, Thornburg KL. Effect of in utero and early-life
conditions on adult health and disease. The New England journal of medicine 2008; 359(1):
61-73.
34. Levine TA, Grunau RE, McAuliffe FM, Pinnamaneni R, Foran A, Alderdice FA. Early
Childhood Neurodevelopment After Intrauterine Growth Restriction: A Systematic
Review. Pediatrics 2015; 135(1): 126-41.
35. Wang X, Zhu J, Guo C, et al. Growth of infants and young children born small for
gestational age: growth restriction accompanied by overweight. Journal of International
Medical Research; 0(0): 0300060518779305.
36. Institute of Medicine of the National Academies. Preterm Birth: Causes, Consequences,
and Prevention. Washington DC: National Academy of Sciences; 2007.
37. Been JV, Lugtenberg MJ, Smets E, et al. Preterm Birth and Childhood Wheezing
Disorders: A Systematic Review and Meta-Analysis. PLoS Med 2014; 11(1): e1001596.
93
38. Colin AA, McEvoy C, Castile RG. Respiratory morbidity and lung function in preterm
infants of 32 to 36 weeks' gestational age. Pediatrics 2010; 126(1): 115-28.
39. Barker DJP. The origins of the developmental origins theory. Journal of Internal Medicine
2007; 261(5): 412-7.
40. Wadhwa PD, Buss C, Entringer S, Swanson JM. Developmental Origins of Health and
Disease: Brief History of the Approach and Current Focus on Epigenetic Mechanisms.
Seminars in reproductive medicine 2009; 27(5): 358-68.
41. Pisaneschi S, Boldrini A, Genazzani A, Coceani F, Simoncini T. Feto-placental vascular
dysfunction as a prenatal determinant of adult cardiovascular disease. Intern Emerg Med
2013; 8(1): 41-5.
42. Hall JG. Review and hypothesis: Syndromes with severe intrauterine growth restriction
and very short stature—Are they related to the epigenetic mechanism(s) of fetal survival
involved in the developmental origins of adult health and disease? American Journal of
Medical Genetics Part A 2010; 152A(2): 512-27.
43. Rosa MJ, eacute. Prenatal exposure to PM2.5 and birth weight: A pooled analysis from
three North American longitudinal pregnancy cohort studies. Environment international
2017; 107: 173-80.
44. Li X, Huang S, Jiao A, et al. Association between ambient fine particulate matter and
preterm birth or term low birth weight: An updated systematic review and meta-analysis.
Environmental Pollution 2017; 227: 596-605.
45. Liu C, Sun J, Liu Y, et al. Different exposure levels of fine particulate matter and preterm
birth: a meta-analysis based on cohort studies. Environ Sci Pollut Res 2017; 24(22): 17976-
84.
46. Siddika N. Prenatal ambient air pollution exposure and the risk of stillbirth: systematic
review and meta-analysis of the empirical evidence. Occupational and environmental
medicine (London, England) 2016; 73(9): 573-81.
47. Xiao Q, Chen H, Strickland MJ, et al. Associations between birth outcomes and maternal
PM2.5 exposure in Shanghai: A comparison of three exposure assessment approaches.
Environment International 2018; 117: 226-36.
48. Hyder A, Lee HJ, Ebisu K, Koutrakis P, Belanger K, Bell ML. PM2.5 exposure and birth
outcomes: use of satellite- and monitor-based data. Epidemiology 2014; 25(1): 58-67.
49. Kloog I, Melly SJ, Ridgway WL, Coull BA, Schwartz J. Using new satellite based exposure
methods to study the association between pregnancy PM2.5 exposure, premature birth and
birth weight in Massachusetts. Environmental Health: A Global Access Science Source
2012; 11: 40.
94
50. Järup L, Åkesson A. Current status of cadmium as an environmental health problem.
Toxicology and Applied Pharmacology 2009; 238(3): 201-8.
51. Agency for Toxic Substances and Disease Registry. Toxicological profile for cadmium,
2012.
52. Lee W, Lee S, Roh J, Won J-U, Yoon J-H. The Association between Involuntary Smoking
Exposure with Urine Cotinine Level and Blood Cadmium Level in General Non-Smoking
Populations. Journal of Korean Medical Science 2017; 32(4): 568-75.
53. Garner R, Levallois P. Cadmium levels and sources of exposure among Canadian adults.
Health Reports 2016; 27(2): 10-8.
54. Berglund M, Larsson K, Grandér M, et al. Exposure determinants of cadmium in European
mothers and their children. Environmental Research 2015; 141: 69-76.
55. Barn P, Gombojav E, Ochir C, et al. The effect of portable HEPA filter air cleaners on
indoor PM2.5 concentrations and second hand tobacco smoke exposure among pregnant
women in Ulaanbaatar, Mongolia: The UGAAR randomized controlled trial. Science of
The Total Environment 2018; 615: 1379-89.
56. World Health Organization. Air quality guidelines for Europe, 2000.
57. The World Bank. Electricity production from coal sources (% of total). 2016.
http://data.worldbank.org/indicator/EG.ELC.COAL.ZS/countries?display=default2016).
58. Hanley GE, Janssen PA. Ethnicity-specific birthweight distributions improve identification
of term newborns at risk for short-term morbidity. American journal of obstetrics and
gynecology 2013; 209(5): 428.e1-.e6.
59. Sabra S, Malmqvist E, Saborit A, Gratacós E, Gomez Roig MD. Heavy metals exposure
levels and their correlation with different clinical forms of fetal growth restriction. PLoS
ONE 2017; 12(10): e0185645.
60. Vidal AC, Semenova V, Darrah T, et al. Maternal cadmium, iron and zinc levels, DNA
methylation and birth weight. BMC Pharmacology and Toxicology 2015; 16(1): 20.
61. Al-Saleh I, Shinwari N, Mashhour A, Rabah A. Birth outcome measures and maternal
exposure to heavy metals (lead, cadmium and mercury) in Saudi Arabian population. Int J
Hyg Environ Health 2014; 217(2-3): 205-18.
62. Johnston JE, Valentiner E, Maxson P, Miranda ML, Fry RC. Maternal cadmium levels
during pregnancy associated with lower birth weight in infants in a North Carolina cohort.
PLoS One 2014; 9(10): e109661.
95
63. Sun H, Chen W, Wang D, Jin Y, Chen X, Xu Y. The effects of prenatal exposure to low-
level cadmium, lead and selenium on birth outcomes. Chemosphere 2014; 108: 33-9.
64. Ikeh-Tawari EP, Anetor JI, Charles-Davies MA. Cadmium Level in Pregnancy, Influence
on Neonatal Birth Weight and Possible Amelioration by Some Essential Trace Elements.
Toxicology International 2013; 20(1): 108-12.
65. Menai M, Heude B, Slama R, et al. Association between maternal blood cadmium during
pregnancy and birth weight and the risk of fetal growth restriction: The EDEN mother–
child cohort study. Reproductive Toxicology 2012; 34(4): 622-7.
66. Lin C-MCM. Does prenatal cadmium exposure affect fetal and child growth? Occupational
and environmental medicine (London, England) 2011; 68(9): 641-6.
67. Nishijo M, Tawara K, Honda R, Nakagawa H, Tanebe K, Saito S. Relationship Between
Newborn Size and Mother's Blood Cadmium Levels, Toyama, Japan. Archives of
Environmental Health: An International Journal 2004; 59(1): 22-5.
68. Zhang Y-LYL. Effect of environmental exposure to cadmium on pregnancy outcome and
fetal growth: a study on healthy pregnant women in China. Journal of environmental
science and health Part A, Toxic/hazardous substances & environmental engineering
2004; 39(9): 2507-15.
69. Salpietro CD, Gangemi S, Minciullo PL, et al. Cadmium concentration in maternal and
cord blood and infant birth weight: a study on healthy non-smoking women. Journal of
perinatal medicine 2002; 30(5): 395-9.
70. Odland J, Nieboer E, Romanova N, Thomassen Y, Lund E. Blood lead and cadmium and
birth weight among sub-arctic and arctic populations of Norway and Russia. Acta
Obstetricia et Gynecologica Scandinavica 1999; 78(10): 852-60.
71. Yang J, Huo W, Zhang B, et al. Maternal urinary cadmium concentrations in relation to
preterm birth in the Healthy Baby Cohort Study in China. Environment International 2016;
94(Supplement C): 300-6.
72. Shirai S, Suzuki Y, Yoshinaga J, Mizumoto Y. Maternal exposure to low-level heavy
metals during pregnancy and birth size. Journal of Environmental Science and Health, Part
A 2010; 45(11): 1468-74.
73. Everson TM, Armstrong DA, Jackson BP, Green BB, Karagas MR, Marsit CJ. Maternal
cadmium, placental PCDHAC1, and fetal development. Reproductive toxicology
(Elmsford, NY) 2016; 65: 263-71.
74. Jin L, Liu J, Ye B, Ren A. Concentrations of selected heavy metals in maternal blood and
associated factors in rural areas in Shanxi Province, China. Environment International
2014; 66(0): 157-64.
96
75. Zhang Y, Cao S, Xu X, et al. Metals compositions of indoor PM2.5, health risk assessment,
and birth outcomes in Lanzhou, China. Environmental monitoring and assessment 2016;
188(6): 325.
76. Kannan S, Misra DP, Dvonch JT, Krishnakumar A. Exposures to Airborne Particulate
Matter and Adverse Perinatal Outcomes: A Biologically Plausible Mechanistic Framework
for Exploring Potential Effect Modification by Nutrition. Environmental Health
Perspectives 2006; 114(11): 1636-42.
77. Erickson AC, Arbour L. The Shared Pathoetiological Effects of Particulate Air Pollution
and the Social Environment on Fetal-Placental Development. Journal of Environmental
and Public Health 2014; 2014: 20.
78. Proietti E, Roosli M, Frey U, Latzin P. Air pollution during pregnancy and neonatal
outcome: a review. J Aerosol Med Pulm Drug Deliv 2013; 26(1): 9-23. doi:
10.1089/jamp.2011.0932. Epub 2012 Aug 2.
79. Ewa B, Danuta M-Š. Polycyclic aromatic hydrocarbons and PAH-related DNA adducts.
Journal of Applied Genetics 2017; 58(3): 321-30.
80. Furness DLF, Dekker GA, Roberts CT. DNA damage and health in pregnancy. Journal of
reproductive immunology 2011; 89(2): 153-62.
81. Lucie S, Petr P. The Function of Cytochrome P450 1A1 Enzyme (CYP1A1) and Aryl
Hydrocarbon Receptor (AhR) in the Placenta. Current Pharmaceutical Biotechnology
2011; 12(5): 715-30.
82. Kippler M, Hoque AMW, Raqib R, Öhrvik H, Ekström E-C, Vahter M. Accumulation of
cadmium in human placenta interacts with the transport of micronutrients to the fetus.
Toxicology Letters 2010; 192(2): 162-8.
83. Hutcheon JA, Zhang X, Cnattingius S, Kramer MS, Platt RW. Customised birthweight
percentiles: does adjusting for maternal characteristics matter? BJOG: An International
Journal of Obstetrics & Gynaecology 2008; 115(11): 1397-404.
84. Martinez SR, Gay MS, Zhang L. Epigenetic mechanisms in heart development and disease.
Drug Discovery Today 2015; 20(7): 799-811.
85. Visentin S, Grumolato F, Nardelli GB, Di Camillo B, Grisan E, Cosmi E. Early origins of
adult disease: Low birth weight and vascular remodeling. Atherosclerosis 2014; 237(2):
391-9.
86. Henderson DE, Milford JB, Miller SL. Prescribed burns and wildfires in Colorado: impacts
of mitigation measures on indoor air particulate matter. Journal of the Air & Waste
Management Association (1995) 2005; 55(10): 1516-26.
97
87. Zhan Y, Johnson K, Norris C, et al. The influence of air cleaners on indoor particulate
matter components and oxidative potential in residential households in Beijing. Science of
The Total Environment 2018; 626: 507-18.
88. Shao D, Du Y, Liu S, et al. Cardiorespiratory responses of air filtration: A randomized
crossover intervention trial in seniors living in Beijing: Beijing Indoor Air Purifier StudY,
BIAPSY. Science of The Total Environment 2017; 603-604: 541-9.
89. Kajbafzadeh M, Brauer M, Karlen B, Carlsten C, van Eeden S, Allen RW. The impacts of
traffic-related and woodsmoke particulate matter on measures of cardiovascular health: a
HEPA filter intervention study. Occupational and Environmental Medicine 2015; 72(6):
394-400.
90. Batterman S, Du L, Mentz G, et al. Particulate matter concentrations in residences: an
intervention study evaluating stand-alone filters and air conditioners. Indoor Air 2012;
22(3): 235-52.
91. Allen RW, Carlsten C, Karlen B, et al. An air filter intervention study of endothelial
function among healthy adults in a woodsmoke-impacted community. American Journal
Of Respiratory & Critical Care Medicine 2011; 183(9): 1222-30.
92. Butz AM, Matsui EC, Breysse P, et al. A randomized trial of air cleaners and a health coach
to improve indoor air quality for inner-city children with asthma and secondhand smoke
exposure. Archives of Pediatrics & Adolescent Medicine 2011; 165(8): 741-8.
93. Lanphear BP, Hornung RW, Khoury J, Yolton K, Lierl M, Kalkbrenner A. Effects of
HEPA air cleaners on unscheduled asthma visits and asthma symptoms for children
exposed to secondhand tobacco smoke. Pediatrics 2011; 127(1): 93-101.
94. Barn P, Larson T, Noullett M, Kennedy S, Copes R, Brauer M. Infiltration of forest fire
and residential wood smoke: an evaluation of air cleaner effectiveness. Journal of exposure
science & environmental epidemiology 2008; 18(5): 503-11.
95. Brauner EV, Forchhammer L, Moller P, et al. Indoor particles affect vascular function in
the aged: An air filtration-based intervention study. Am J Respir Crit Care Med 2008;
177(4): 419-25.
96. Weichenthal S, Mallach G, Kulka R, et al. A randomized double-blind crossover study of
indoor air filtration and acute changes in cardiorespiratory health in a First Nations
community. Indoor Air 2013; 23(3): 175-84.
97. Ochir C, Smith KR. Air pollution and health in Ulaanbaatar, 2014.
98. Guttikunda S. Urban air pollution analysis for Ulaanbaatar, The World Bank Consultation
Report. Washington, DC, 2007.
98
99. The World Bank. Mongolia, heating in poor, peri-urban ger areas of Ulaanbaatar, 2009.
100. Hill LD, Edwards R, Turner JR, et al. Health assessment of future PM2.5 exposures from
indoor, outdoor, and secondhand tobacco smoke concentrations under alternative policy
pathways in Ulaanbaatar, Mongolia. PLOS ONE 2017; 12(10): e0186834.
101. The World Bank Group. Mongolia - Heating stove market trends in poor, peri-urban ger
areas of Ulaanbaatar and selected markets outside Ulaanbaatar : stocktaking report of the
Mongolia clean stoves initiative 2013.
102. Kamata T, Reichert J, Tsevegmid T, Kim Y, Sedgewick B. Mongolia - Enhancing policies
and practices for ger area development in Ulaanbaatar Washington, DC: World Bank,
2010.
103. Institute for Health Metrics and Evaluation. Global Burden of Disease Study 2015 (GBD
2015) Smoking Prevalence 1980-2015. 2017. http://ghdx.healthdata.org/record/global-
burden-disease-study-2015-gbd-2015-smoking-prevalence-1980-2015.
104. Forouzanfar MH, Afshin A, Alexander LT, et al. Global, regional, and national
comparative risk assessment of 79 behavioural, environmental and occupational, and
metabolic risks or clusters of risks, 1990–2015: a systematic analysis for the Global Burden
of Disease Study 2015. The Lancet 2016; 388(10053): 1659-724.
105. Brauer M, Freedman G, Frostad J, et al. Ambient Air Pollution Exposure Estimation for
the Global Burden of Disease 2013. Environmental Science & Technology 2016; 50(1):
79-88.
106. Stieb DM. Estimated public health impacts of changes in concentrations of fine particle air
pollution in Canada, 2000 to 2011. Canadian journal of public health 2015; 106(6): e362-
8.
107. Pope CA, Ezzati M, Dockery DW. Fine Particulate Air Pollution and US County Life
Expectancies. The New England journal of medicine 2009; 360(4): 376-86.
108. Gauderman WJ, Urman R, Avol E, et al. Association of Improved Air Quality with Lung
Development in Children. The New England journal of medicine 2015; 372(10): 905-13.
109. Fenger J. Urban air quality. Atmospheric Environment 1999; 33(29): 4877-900.
110. Allen RW, Adar SD, Avol E, et al. Modeling the Residential Infiltration of Outdoor
PM(2.5) in the Multi-Ethnic Study of Atherosclerosis and Air Pollution (MESA Air).
Environmental Health Perspectives 2012; 120(6): 824-30.
99
111. Clark NA, Allen RW, Hystad P, et al. Exploring Variation and Predictors of Residential
Fine Particulate Matter Infiltration. International Journal of Environmental Research and
Public Health 2010; 7(8): 3211-24.
112. Xu CC. Investigation and modeling of the residential infiltration of fine particulate matter
in Beijing. Journal of the Air & Waste Management Association (1995) 2016: 1-8.
113. Leech JA, Nelson WC, Burnett RT, Aaron S, Raizenne ME. It's about time: A comparison
of Canadian and American time-activity patterns[dagger]. Journal of exposure analysis
and environmental epidemiology 2002; 12(6): 427-32.
114. Fisk WJ, Chan WR. Effectiveness and cost of reducing particle-related mortality with
particle filtration. Indoor Air 2017; [Epub ahead of print].
115. McNamara ML, Thornburg J, Semmens EO, Ward TJ, Noonan CW. Reducing indoor air
pollutants with air filtration units in wood stove homes. Science of The Total Environment
2017; 592: 488-94.
116. Chen R, Zhao A, Chen H, et al. Cardiopulmonary Benefits of Reducing Indoor Particles of
Outdoor Origin: a Randomized Double-Blind Crossover Trial of Air Purifiers. Journal of
the American College of Cardiology 2015; 65(21): 2279-87.
117. Ward TJ, Semmens EO, Weiler E, Harrar S, Noonan CW. Efficacy of interventions
targeting household air pollution from residential wood stoves. J Expos Sci Environ
Epidemiol 2017; 27(1): 64-71.
118. Reitsma MB, Fullman N, Ng M, et al. Smoking prevalence and attributable disease burden
in 195 countries and territories, 1990-2015: a systematic analysis from the Global Burden
of Disease Study 2015. The Lancet 2017.
119. Fisk WJ. Health benefits of particle filtration. Indoor Air 2013; 23(5): 357-68.
120. Mongolian Statistical Information Service. Statistical information. 2016.
http://www.1212.mn/en/.
121. Song D. Contents and Occurrence of Cadmium in the Coals from Guizhou Province, China.
Annals of the New York Academy of Sciences 2008; 1140: 274-81.
122. Allen RW, Gombojav E, Barkhasragchaa B, et al. An assessment of air pollution and its
attributable mortality in Ulaanbaatar, Mongolia. Air quality, atmosphere, & health 2013;
6(1): 137-50.
123. Northcross AL, Edwards RJ, Johnson MA, et al. A low-cost particle counter as a realtime
fine-particle mass monitor. Environmental Science: Processes & Impacts 2013; 15(2):
433-9.
100
124. Semple S, Ibrahim AE, Apsley A, Steiner M, Turner S. Using a new, low-cost air quality
sensor to quantify second-hand smoke (SHS) levels in homes. Tobacco control 2015;
24(2): 153-8.
125. Steinle S, Reis S, Sabel CE, et al. Personal exposure monitoring of PM2.5 in indoor and
outdoor microenvironments. Science of The Total Environment 2015; 508: 383-94.
126. Semple S, Apsley A, MacCalman L. An inexpensive particle monitor for smoker behaviour
modification in homes. Tobacco control 2013; 22(5): 295-8.
127. Klepeis NE, Hughes SC, Edwards RD, et al. Promoting smoke-free homes: a novel
behavioral intervention using real-time audio-visual feedback on airborne particle levels.
PLoS One 2013; 8(8): e73251.
128. United States Environmental Protection Agency. Evaluation of field-deployed low cost PM
sensors, 2014.
129. Palmer CD, Lewis Jr ME, Geraghty CM, Barbosa Jr F, Parsons PJ. Determination of lead,
cadmium and mercury in blood for assessment of environmental exposure: A comparison
between inductively coupled plasma–mass spectrometry and atomic absorption
spectrometry. Spectrochimica Acta Part B: Atomic Spectroscopy 2006; 61(8): 980-90.
130. Hornung RW, Reed LD. Estimation of Average Concentration in the Presence of
Nondetectable Values. Applied Occupational and Environmental Hygiene 1990; 5(1): 46-
51.
131. Al-Delaimy WK. Hair as a biomarker for exposure to tobacco smoke. Tobacco control
2002; 11(3): 176-82.
132. Bernhoft RA. Cadmium Toxicity and Treatment. The Scientific World Journal 2013; 2013:
7.
133. International Agency for Research on Cancer. Cadmium and cadmium compounds, 2012.
134. Garner R, Levallois P. Cadmium levels and sources of exposure among Canadian adults.
Health Rep 2016; 27(2): 10-8.
135. Edwards SE, Maxson P, Miranda ML, Fry RC. Cadmium levels in a North Carolina cohort:
Identifying risk factors for elevated levels during pregnancy. Journal of exposure science
& environmental epidemiology 2015; 25(4): 427-32.
136. Hinwood AL, Callan AC, Ramalingam M, et al. Cadmium, lead and mercury exposure in
non smoking pregnant women. Environ Res 2013; 126: 118-24.
101
137. Hansen SS. Changes in maternal blood concentrations of selected essential and toxic
elements during and after pregnancy. Journal of environmental monitoring 2011; 13(8):
2143-52.
138. Taylor CM, Golding J, Emond AM. Lead, cadmium and mercury levels in pregnancy: the
need for international consensus on levels of concern. Journal of Developmental Origins
of Health and Disease 2014; 5(01): 16-30.
139. Yoo SH, Paek YJ, Kim SS, et al. Hair nicotine levels in non-smoking pregnant women
whose spouses smoke outside of the home. Tobacco control 2010; 19(4): 318-24.
140. Seong MW, Hwang JH, Moon JS, et al. Neonatal hair nicotine levels and fetal exposure to
paternal smoking at home. American journal of epidemiology 2008; 168(10): 1140-4.
141. California Environmental Protection Agency Air Resources Board. Air cleaning devices
for the home. 2014. https://www.arb.ca.gov/research/indoor/acdsumm.pdf.
142. Fisk WJ, Chan WR. Health benefits and costs of filtration interventions that reduce indoor
exposure to PM2.5 during wildfires. Indoor Air 2016; 27(1): 191-2014.
143. GBD Risk Factors Collaborators. Global, regional, and national comparative risk
assessment of 84 behavioural, environmental and occupational, and metabolic risks or
clusters of risks, 1990–2016: a systematic analysis for the Global Burden of Disease Study
2016. Lancet (London, England) 2017; 390(10100): 1345-422.
144. Thompson LM, Bruce N, Eskenazi B, Diaz A, Pope D, Smith KR. Impact of reduced
maternal exposures to wood smoke from an introduced chimney stove on newborn birth
weight in rural Guatemala. Environ Health Perspect 2011; 119(10): 1489-94.
145. Kiserud T, Piaggio G, Carroli G, et al. The World Health Organization Fetal Growth
Charts: A Multinational Longitudinal Study of Ultrasound Biometric Measurements and
Estimated Fetal Weight. PLoS Medicine 2017; 14(1): e1002220.
146. Morello-Frosch R, Jesdale BM, Sadd JL, Pastor M. Ambient air pollution exposure and
full-term birth weight in California. Environ Health 2010; 9: 44.
147. Bell ML, Ebisu K, Belanger K. Ambient air pollution and low birth weight in Connecticut
and Massachusetts. Environ Health Perspect 2007; 115.
148. Chen C, Zhao B. Review of relationship between indoor and outdoor particles: I/O ratio,
infiltration factor and penetration factor. Atmospheric Environment 2011; 45(2): 275-88.
149. Nethery E, Brauer M, Janssen P. Time-activity patterns of pregnant women and changes
during the course of pregnancy. Journal of exposure science & environmental
epidemiology 2009; 19(3): 317-24.
102
150. Mongolian Statistical Information Service. Monthly average wages and salaries, by aimag
and the Capital, sex, quarter. 2017.
http://www.1212.mn/tables.aspx?tbl_id=DT_NSO_0400_021V1&SOUM_select_all=0&
SOUMSingleSelect=_511&Gender_select_all=0&GenderSingleSelect=_1&Year_select_
all=0&YearSingleSelect=_2016_2015_2012_2014&viewtype=table.
151. Savitz DA, Bobb JF, Carr JL, et al. Ambient fine particulate matter, nitrogen dioxide, and
term birth weight in New York, New York. American journal of epidemiology 2014;
179(4): 457-66.
152. Rich DQ, Liu K, Zhang J, et al. Differences in Birth Weight Associated with the 2008
Beijing Olympic Air Pollution Reduction: Results from a Natural Experiment. Environ
Health Perspect 2015.
153. Clemens T, Turner S, Dibben C. Maternal exposure to ambient air pollution and fetal
growth in North-East Scotland: A population-based study using routine ultrasound scans.
Environment International 2017; 107: 216-26.
154. van den Hooven EH, Pierik FH, de Kluizenaar Y, et al. Air pollution exposure during
pregnancy, ultrasound measures of fetal growth, and adverse birth outcomes: a prospective
cohort study. Environ Health Perspect 2012; 120(1): 150-6.
155. Brook RD, Rajagopalan S, Pope CA, 3rd, et al. Particulate matter air pollution and
cardiovascular disease: An update to the scientific statement from the American Heart
Association. Circulation 2010; 121(21): 2331-78.
156. Alexander D, Northcross A, Wilson N, et al. Randomized Controlled Ethanol Cookstove
Intervention and Blood Pressure in Pregnant Nigerian Women. American Journal of
Respiratory and Critical Care Medicine 2017; 195(12): 1629-39.
157. Imdad A, Bhutta ZA. Maternal Nutrition and Birth Outcomes: Effect of Balanced Protein-
Energy Supplementation. Paediatric and perinatal epidemiology 2012; 26: 178-90.
158. Gresham E, Byles JE, Bisquera A, Hure AJ. Effects of dietary interventions on neonatal
and infant outcomes: a systematic review and meta-analysis. The American Journal of
Clinical Nutrition 2014; 100(5): 1298-321.
159. Vaivada T, Gaffey MF, Das JK, Bhutta ZA. Evidence-based interventions for improvement
of maternal and child nutrition in low-income settings: what's new? Current Opinion in
Clinical Nutrition & Metabolic Care 2017; 20(3): 204-10.
160. Bhutta ZA, Das JK, Rizvi A, et al. Evidence-based interventions for improvement of
maternal and child nutrition: what can be done and at what cost? The Lancet 2013;
382(9890): 452-77.
103
161. Pineles BL, Park E, Samet JM. Systematic Review and Meta-Analysis of Miscarriage and
Maternal Exposure to Tobacco Smoke During Pregnancy. American journal of
epidemiology 2014; 179(7): 807-23.
162. Enkhmaa D, Warburton N, Javzandulam B, et al. Seasonal ambient air pollution correlates
strongly with spontaneous abortion in Mongolia. BMC pregnancy and childbirth 2014; 14:
146.
163. Pereira LA, Loomis D, Conceição GM, et al. Association between air pollution and
intrauterine mortality in São Paulo, Brazil. Environmental Health Perspectives 1998;
106(6): 325-9.
164. Hou HY, Wang D, Zou XP, Yang ZH, Li T-C, Chen YQ. Does ambient air pollutants
increase the risk of fetal loss? A case–control study. Arch Gynecol Obstet 2014; 289(2):
285-91.
165. Burton GJ, Jauniaux E. Oxidative stress. Best Pract Res Clin Obstet Gynaecol 2011; 25(3):
287-99.
166. El-Mohandes AAE, Kiely M, Blake SM, Gantz MG, El-Khorazaty MN. An Intervention
to Reduce Environmental Tobacco Smoke Exposure Improves Pregnancy Outcomes.
Pediatrics 2010; 125(4): 721-8.
167. Khong Y, Brosens I. Defective deep placentation. Best Practice & Research Clinical
Obstetrics & Gynaecology 2011; 25(3): 301-11.
168. Mifsud W, Sebire NJ. Placental pathology in early-onset and late-onset fetal growth
restriction. Fetal diagnosis and therapy 2014; 36(2): 117-28.
169. Nieuwenhuijsen MJ, Ristovska G, Dadvand P. WHO Environmental Noise Guidelines for
the European Region: A Systematic Review on Environmental Noise and Adverse Birth
Outcomes. International Journal of Environmental Research and Public Health 2017;
14(10): 1252.
170. Shao L, Hu Y, Shen R, et al. Seasonal variation of particle-induced oxidative potential of
airborne particulate matter in Beijing. Science of The Total Environment 2017; 579: 1152-
60.
171. Tian H, Cheng K, Wang Y, et al. Temporal and spatial variation characteristics of
atmospheric emissions of Cd, Cr, and Pb from coal in China. Atmospheric Environment
2012; 50: 157-63.
172. Yuchi W, Knudby A, Cowper J, et al. A description of methods for deriving air pollution
land use regression model predictor variables from remote sensing data in Ulaanbaatar,
Mongolia. The Canadian Geographer / Le Géographe canadien 2016; 60(3): 333-45.
104
173. JonØyvind O, Evert N, Natalya R, Yngvar T, Eiliv L. Blood lead and cadmium and birth
weight among sub‐arctic and arctic populations of Norway and Russia. Acta Obstetricia et
Gynecologica Scandinavica 1999; 78(10): 852-60.
174. Taylor CM, Golding J, Emond AM. Adverse effects of maternal lead levels on birth
outcomes in the ALSPAC study: a prospective birth cohort study. BJOG : an international
journal of obstetrics and gynaecology 2015; 122(3): 322-8.
175. Akerstrom M, Barregard L, Lundh T, Sallsten G. The relationship between cadmium in
kidney and cadmium in urine and blood in an environmentally exposed population.
Toxicology and Applied Pharmacology 2013; 268(3): 286-93.
176. Rani A, Kumar A, Lal A, Pant M. Cellular mechanisms of cadmium-induced toxicity: a
review. International Journal of Environmental Health Research 2014; 24(4): 378-99.
177. Liu H, Xia W, Xu S, et al. Cadmium body burden and pregnancy-induced hypertension.
International Journal of Hygiene and Environmental Health 2018; 221(2): 246-51.
178. Yang K, Julan L, Rubio F, Sharma A, Guan H. Cadmium reduces 11β-hydroxysteroid
dehydrogenase type 2 activity and expression in human placental trophoblast cells.
American Journal of Physiology-Endocrinology and Metabolism 2006; 290(1): E135-E42.
179. Basu R, Harris M, Sie L, Malig B, Broadwin R, Green R. Effects of fine particulate matter
and its constituents on low birth weight among full-term infants in California.
Environmental Research 2014; 128: 42-51.
180. Hao Y, Strosnider H, Balluz L, Qualters JR. Geographic Variation in the Association
between Ambient Fine Particulate Matter (PM) and Term Low Birth Weight in the United
States. Environ Health Perspect 2015.
181. Lavigne É, Burnett R, Stieb D, et al. Fine Particulate Air Pollution and Adverse Birth
Outcomes: Effect Modification by Regional Nonvolatile Oxidative Potential.
Environmental health perspectives 2018; 126(7).
182. Ayres JG, Borm P, Cassee FR, et al. Evaluating the Toxicity of Airborne Particulate Matter
and Nanoparticles by Measuring Oxidative Stress Potential—A Workshop Report and
Consensus Statement. Inhalation Toxicology 2008; 20(1): 75-99.
183. Künzli N, Mudway IS, Götschi T, et al. Comparison of Oxidative Properties, Light
Absorbance, and Total and Elemental Mass Concentration of Ambient PM(2.5) Collected
at 20 European Sites. Environmental Health Perspectives 2006; 114(5): 684-90.
105
184. Jedrychowski WA, Majewska R, Spengler JD, Camann D, Roen EL, Perera FP. Prenatal
exposure to fine particles and polycyclic aromatic hydrocarbons and birth outcomes: a two-
pollutant approach. International Archives of Occupational and Environmental Health
2017; 90(3): 255-64.
185. Choi H, Rauh V, Garfinkel R, Tu Y, Perera FP. Prenatal Exposure to Airborne Polycyclic
Aromatic Hydrocarbons and Risk of Intrauterine Growth Restriction. Environmental
Health Perspectives 2008; 116(5): 658-65.
186. Jung SY, Kim S, Lee K, et al. Association between secondhand smoke exposure and blood
lead and cadmium concentration in community dwelling women: the fifth Korea National
Health and Nutrition Examination Survey (2010–2012). BMJ open 2015; 5(7).
187. Sun H, Chen W, Wang D, Jin Y, Chen X, Xu Y. The effects of prenatal exposure to low-
level cadmium, lead and selenium on birth outcomes. Chemosphere 2014; 108: 33-9.
188. Cheng K, Tian HZ, Zhao D, et al. Atmospheric emission inventory of cadmium from
anthropogenic sources. International Journal of Environmental Science and Technology
2014; 11(3): 605-16.
189. Kragie SX, Ryan PB, Bergin MH, Wang S. Airborne trace metals from coal combustion in
Beijing. Air Quality, Atmosphere & Health 2013; 6(1): 157-65.
190. Kim ND, Fergusson JE. The concentrations, distribution and sources of cadmium, copper,
lead and zinc in the atmosphere of an urban environment. Science of The Total Environment
1994; 144(1): 179-89.
191. Bray C, Battye W, Uttamang P, Pillai P, Aneja V. Characterization of Particulate Matter
(PM2.5 and PM10) Relating to a Coal Power Plant in the Boroughs of Springdale and
Cheswick, PA. Atmosphere 2017; 8(10): 186.
192. Downward GS, Hu W, Large D, et al. Heterogeneity in coal composition and implications
for lung cancer risk in Xuanwei and Fuyuan counties, China. Environment International
2014; 68: 94-104.
193. Shi J, Huang W, Chen P, Tang S, Chen X. Concentration and Distribution of Cadmium in
Coals of China. Minerals 2018; 8(2): 48.
194. Cheng S, Liu G, Liu Y, Wu D. Cadmium in Chinese coals: Abundance, distribution,
occurrence, and environmental effects. Human and Ecological Risk Assessment: An
International Journal 2018: 1-21.
195. Adams SV, Newcomb PA. Cadmium blood and urine concentrations as measures of
exposure: NHANES 1999-2010. J Expos Sci Environ Epidemiol 2014; 24(2): 163-70.
106
196. Al-Rmalli SW, Jenkins RO, Haris PI. Dietary Intake of Cadmium from Bangladeshi Foods.
Journal of Food Science 2012; 77(1): T26-T33.
197. Fløtre CH, Varsi K, Helm T, Bolann B, Bjørke-Monsen A-L. Predictors of mercury, lead,
cadmium and antimony status in Norwegian never-pregnant women of fertile age. PLOS
ONE 2017; 12(12): e0189169.
198. Ganmaa D, Rich-Edwards JW, Frazier LA, et al. A comparison of migrants to, and women
born in, urban Mongolia: demographic, reproductive, anthropometric and lifestyle
characteristics. International health 2013; 5(4): 244-50.
199. Arbuckle TE, Liang CL, Morisset A-S, et al. Maternal and fetal exposure to cadmium, lead,
manganese and mercury: The MIREC study. Chemosphere 2016; 163: 270-82.
200. Salam RA, Das JK, Bhutta ZA. Impact of intrauterine growth restriction on long-term
health. Current opinion in clinical nutrition and metabolic care 2014; 17(3): 249-54.
201. Fleischer NL, Merialdi M, van Donkelaar A, et al. Outdoor air pollution, preterm birth, and
low birth weight: analysis of the world health organization global survey on maternal and
perinatal health. Environ Health Perspect 2014; 122(4): 425-30.
202. Zhao N, Qiu J, Zhang Y, et al. Ambient air pollutant PM10 and risk of preterm birth in
Lanzhou, China. Environ Int 2015; 76: 71-7.
203. Zhao Q, Liang Z, Tao S, Zhu J, Du Y. Effects of air pollution on neonatal prematurity in
Guangzhou of China: a time-series study. Environ Health 2011; 10: 2.
204. Jacobs M, Zhang G, Chen S, et al. The association between ambient air pollution and
selected adverse pregnancy outcomes in China: A systematic review. Science of The Total
Environment 2017; 579(Supplement C): 1179-92.
205. Kerimray A, Rojas-Solórzano L, Amouei Torkmahalleh M, Hopke PK, Ó Gallachóir BP.
Coal use for residential heating: Patterns, health implications and lessons learned. Energy
for Sustainable Development 2017; 40(Supplement C): 19-30.
206. Thurston GD, Burnett RT, Turner MC, et al. Ischemic Heart Disease Mortality and Long-
Term Exposure to Source-Related Components of U.S. Fine Particle Air Pollution.
Environmental Health Perspectives 2016; 124(6): 785-94.
207. Laden F, Neas LM, Dockery DW, Schwartz J. Association of fine particulate matter from
different sources with daily mortality in six U.S. cities. Environmental Health Perspectives
2000; 108(10): 941-7.
107
208. Ito K, Christensen WF, Eatough DJ, et al. PM source apportionment and health effects: 2.
An investigation of intermethod variability in associations between source-apportioned
fine particle mass and daily mortality in Washington, DC. Journal Of Exposure Science
And Environmental Epidemiology 2005; 16: 300.
209. Agarwal P, Rajadurai VS, Yap F, et al. Comparison of customized and cohort-based
birthweight standards in identification of growth-restricted infants in GUSTO cohort study.
The Journal of Maternal-Fetal & Neonatal Medicine 2015: 1-4.
210. Hemming K, Bonellie S, Hutton JL. Fetal growth and birthweight standards as screening
tools: methods for evaluating performance. BJOG: An International Journal of Obstetrics
& Gynaecology 2011; 118(12): 1477-83.
211. Zheng L, Liu G, Chou C-L. The distribution, occurrence and environmental effect of
mercury in Chinese coals. Science of The Total Environment 2007; 384(1–3): 374-83.
212. Fang T, Liu G, Zhou C, Sun R, Chen J, Wu D. Lead in Chinese coals: distribution, modes
of occurrence, and environmental effects. Environ Geochem Health 2013; 36(3): 563-81.
213. Amegah AK, Quansah R, Jaakkola JJ. Household air pollution from solid fuel use and risk
of adverse pregnancy outcomes: a systematic review and meta-analysis of the empirical
evidence. PLoS One 2014; 9(12): e113920.
214. Ha S, Hu H, Roth J, Kan H, Xu X. Associations Between Residential Proximity to Power
Plants and Adverse Birth Outcomes. American journal of epidemiology 2015; 182(3): 215-
24.
215. Landrigan PJ, Fuller R, Acosta NJR, et al. The Lancet Commission on pollution and health.
Lancet 2018; 391(10119): 462-512.
216. BC Centre for Disease Control. Guidance for BC public health decision makers during
wildfire smoke events, 2014.
217. Barn PK, Elliott CT, Allen RW, Kosatsky T, Rideout K, Henderson SB. Portable air
cleaners should be at the forefront of the public health response to landscape fire smoke.
Environmental Health 2016; 15(1): 116.
218. National Environment Agency. Portable air cleaners. 2018.
https://www.haze.gov.sg/resources/portable-air-cleaners.
219. US Environmental Protection Agency. Wildfire smoke, a guide for public health officials.
2016.
220. Bates DV. The Effects of Air Pollution on Children. Environmental Health Perspectives
1995; 103: 49-53.
108
221. Pope CA, Dockery DW. Health Effects of Fine Particulate Air Pollution: Lines that
Connect. Journal of the Air & Waste Management Association 2006; 56(6): 709-42.
222. US Environmental Protection Agency. The benefits and costs of the Clean Air Act from
1990 to 2010, 2011.
223. Lurmann F, Avol E, Gilliland F. Emissions reduction policies and recent trends in Southern
California’s ambient air quality. Journal of the Air & Waste Management Association
2015; 65(3): 324-35.
224. He G, Fan M, Zhou M. The effect of air pollution on mortality in China: Evidence from
the 2008 Beijing Olympic Games. Journal of Environmental Economics and Management
2016; 79: 18-39.
225. Yixuan Z, Tao X, Qiang Z, et al. Air quality improvements and health benefits from
China’s clean air action since 2013. Environmental Research Letters 2017; 12(11): 114020.
226. Broome RA, Fann N, Cristina TJN, Fulcher C, Duc H, Morgan GG. The health benefits of
reducing air pollution in Sydney, Australia. Environmental Research 2015; 143: 19-25.
227. Millennium Challenge Corporation. Summary: measuring results of the Mongolia energy
and enviroment project stoves subsidies component. 2014.
https://www.mcc.gov/resources/doc/summary-measuring-results-of-mng-eep2016).
228. Woodruff TJ, Zota AR, Schwartz JM. Environmental Chemicals in Pregnant Women in
the United States: NHANES 2003–2004. Environmental Health Perspectives 2011;
119(6): 878-85.
229. Rokoff LB, Rifas-Shiman SL, Coull BA, et al. Cumulative exposure to environmental
pollutants during early pregnancy and reduced fetal growth: the Project Viva cohort.
Environmental Health 2018; 17: 19.
230. Dominici F, Peng RD, Barr CD, Bell ML. Protecting Human Health from Air Pollution:
Shifting from a Single-Pollutant to a Multi-pollutant Approach. Epidemiology (Cambridge,
Mass) 2010; 21(2): 187-94.
231. Smith RB, Fecht D, Gulliver J, et al. Impact of London's road traffic air and noise pollution
on birth weight: retrospective population based cohort study. The BMJ 2017; 359: j5299.
232. Pope CA, Burnett RT, Turner MC, et al. Lung Cancer and Cardiovascular Disease
Mortality Associated with Ambient Air Pollution and Cigarette Smoke: Shape of the
Exposure–Response Relationships. Environmental Health Perspectives 2011; 119(11):
1616-21.
109
233. Pope CA, 3rd, Burnett RT, Krewski D, et al. Cardiovascular mortality and exposure to
airborne fine particulate matter and cigarette smoke: shape of the exposure-response
relationship. Circulation 2009; 120(11): 941-8.
234. Yu H-L, Chien L-C. Short-term population-based non-linear concentration-response
associations between fine particulate matter and respiratory diseases in Taipei (Taiwan): a
spatiotemporal analysis. J Expos Sci Environ Epidemiol 2015.
235. Stieb DM, Chen L, Beckerman BS, et al. Associations of Pregnancy Outcomes and
PM(2.5) in a National Canadian Study. Environmental Health Perspectives 2016; 124(2):
243-9.
110
Appendix A. Summary of studies investigating maternal cadmium exposure and fetal
growth
Table A.1. Summary of studies investigating the relationships between maternal cadmium exposure and fetal growth
Reference
Study setting, sample size,
timing of sample
collection, and health
outcomes of interest
Cadmium concentrations Main findings
Blood
Sabra et al. 20171 Barcelona, Spain
n=178
Samples collected on day
of delivery
IUGRa, small for
gestational age (SGA)
Median (IQR) values by
type of growth restriction:
AGA: 0.04 (0.009) ug/dL
IUGR: 0.05 (0.005) ug/dL
SGA: 0.05 (0.005) ug/dL.
Maternal serum cadmium (Cd) was significantly
higher (p<0.001) in IUGR and SGA fetuses
compared with AGA fetuses (appropriate for
gestational age).
Cheng et al. 20172
Wuhan, China
n=282
Samples collected at 13, 24,
and 35 weeks gestation
Birth weight (BW),
ponderal index (PI), birth
length (BL)
GM (25th, 75th percentile)
by trimester:
1st: 0.51 (0.36, 0.77) ug/g
creatinine;
2nd: 0.59 (0.41, 0.79) ug/g
creatinine;
3rd: 0.61 (0.39, 0.92) ug/g
creatinine
Each log unit increase in first trimester urinary
Cd (ug/g creatinine) was associated with a mean
decrease in BW of 117 g (95% CI: -209, -25 g)
in girls. Cd levels in the first and second
trimesters were also borderline associated with
PI among girls. No significant effects were seen
among boys.
Luo et al. 20173 North Carolina, US
n=275
Samples collected in the
first trimester.
Mean or median
concentrations were not
provided for all
participants, but authors
provided median (IQR)
concentrations by different
Regression models were run for tertiles of Cd
exposure, and analyses were stratified by sex
and smoking. Among boys, BW was
significantly lower among the highest tertile
group compared with the lowest (-812 g, SE =
111
BW
characteristics (e.g. by
maternal age, education,
smoking status, baby’s
sex). Median
concentrations ranged from
0.014 to 0.051 µg/dL.
346, p = 0.02). No significant effects seen
among all births or girls.
The relationship between Cd and BW was
modified by select nutrient combinations (e.g.
iron, folate, calcium).
Taylor et al. 20164 Avon, UK
n=4,191
Samples collected at ~11
weeks gestation.
BW, BL, head
circumference (HC), low
birth weight (LBW)
Median (IQR) was 0.29
(0.14–0.68) µg/L
Cd concentrations were associated with
decreased birthweight (unstandardized B
coefficient -62.7 g, 95% CI: -107.0, -18.4) and
crown–heel length (-0.28 cm, 95% CI: -0.48, -
0.07) per 1 µg/L increase in Cd, in adjusted
regression models. Adjustment variables were:
maternal educational attainment, age, parity,
pre-pregnancy BMI, height, and alcohol intake,
and sex of baby.
Stratification by sex showed adverse effect on
birthweight (-87.1 g, 95% CI: -144.8, -29.4),
head circumference (-0.22 cm, 95% CI: -0.39, -
0.04), and crown–heel length
(-0.44 cm, 95% CI: -0.71, -0.18) in girls only.
Wang et al. 20165 6 cities, China
n=3,254
Samples collected in first
and second trimesters.
SGA
Median (25th and 75th
percentile) was 0.79 (0.57,
1.06) μg/L
Women with cd concentrations > 75th percentile
(1.06 µg/L) during pregnancy had increased
odds of SGA [OR = 1.43 (95% CI: 1.09, 1.88), p
= 0.009], compared to women with
concentrations ≤ 75th percentile, after adjusting
for pre-pregnancy BMI, maternal age, time for
collecting serum, monthly income, and
gravidity.
When stratified by trimester, a higher risk of
SGA was reported for second trimester
concentrations (OR = 1.57; 95% CI: 1.13, 2.19,
p=0.007) for women in higher exposure vs low
112
exposure group. No significant effects seen for
first trimester concentrations.
Bloom et al. 20156 Michigan & Texas, US
N=235
Samples collect pre-
conception
BW, BL, HC, PI,
gestational age (GA)
Mean (SD) was 0.24 (0.14)
μg/L
Cd concentrations were associated with higher
BW, with women in the third tertile having
heavier babies (179g; 25, 332g) compared with
women in the first tertile.
Hu et al. 20157 4 cities, China
N=81
Samples collected on day
of delivery
BW
Median (25th, 75th
percentile) was 0.9 (0.7,
1.2) ng/g.
No associations were found between Cd and
BW.
Rollin et al. 20158 South Africa
n=641
Samples collected on day
of delivery
BW, BL, HC, PI
Geometric mean (95% CI)
was 0.25 μg/L (0.23, 0.27
μg/L)
An association between Cd concentrations and
birth weight percentile was found in girls only (β
= -0.13 for women in the third tertile versus first
tertile of Cd concentration, p = 0.047).
Thomas et al. 20159 10 cities, Canada
n=1,835
Samples collected during
first and third trimesters
SGA
Median (IQR) blood
cadmium level was
0.20 (0.13 – 0.30) µg/L.
Cd concentrations were not associated with risk
of SGA.
113
Vidal et al. 201510 North Carolina, US
n=319
Samples collected at ≤ 12
weeks gestation
BW
Mean or median
concentrations were not
presented for all
participants, but authors
provided means by
different characteristics
(e.g. by sex, age, BMI).
Mean values ranged from
2.56 to 6.48 ng/L.
Cd concentrations were associated with birth
weight (-52g for each log-unit increase; p =
0.03).
Al-Saleh et al.
201411
Al-Kharj, Saudi Arabia
n=1,578
Samples collected at
delivery.
BW, BL, HC, crown-heel
length, Apgar 1-min score,
Apgar 5-min score
Mean (SD) concentration
was 0.99 (0.31) μg/L.
Cd concentrations were significantly higher for
babies who were < 10th percentile for crown-heel
length: OR (95%CI) was 1.644 (1.058, 2.555), p
= 0.027.
Johnston et al.
201412
Durham, North Carolina
N=1,027
Samples collected at
delivery.
BW, SGA, LBW, preterm
birth (PTB), GA, HC, BL
Median (IQR) was 0.40
(0.33) µg/L.
Women in the highest tertile (≥50 µg/L) had
lower birth weight percentile by gestational age
(-6 g, SE: 2.11, p = 0.007) and increased risk of
SGA (OR =1.72, 95% CI: 1.1, 2.68, p = 0.001)
compared with women the lowest tertile
(≤0.28 µg/L).
Sun et al. 201413 Eastern China
n=209
Samples collected in third
trimester
BW, BL
Geometric mean (95% CI)
concentration was 0.48
(0.43, 0.53) µg/L.
Cd was significantly correlated with BW (r=-
0.41; p<0.05).
114
Ikeh-Tawari et al.
201314
Nigeria
n=125
Samples collected at
various trimesters
depending on when
participant was recruited.
BW, BL, HC, LBW
Mean (SD) concentrations
by trimester:
1st: 0.20 (0.10) µmol/L
2nd: 0.21 (0.10) µmol/L
3rd: 0.25 (0.20) µmol/L
Third trimester Cd concentrations were
significantly (p<0.05) correlated with BW (-
0.708), BL (-0.499), and HC (-0.332). Cd
concentrations were also significantly higher
among women who delivered LBW [0.03 (0.10)
µmol/L vs 0.02 (0.10) µmol/L; p=0.02; n=51).
Menai et al. 201215 France
n=901
Samples collected between
24-28 weeks gestation
BW, SGA
Median (IQR)
concentration was 0.8 (0.1-
4.6) µg/L.
Cd concentrations were significantly correlated
with BW in smokers (r= -0.25, p<0.001) but not
in non-smokers (r = 0,03; p=0.398). A dose-
response relationship was observed between
exposure and BW among smokers only.
BW was 204 g lower (p=0.007) among women
in the highest tertile (>1.5 µg/L) compared with
lowest tertile (<1 µg/L). Effects of Cd in the
high exposure group and smoking in pregnancy
had a similar effect on risk of SGA, with ORs of
1.41 (95% CI: 1.00, 1.99) and 1.89 (95% CI:
1.00, 3.58), respectively.
Lin et al. 201116 Taiwan
n=321
Samples collected at
delivery.
BL, HC, PTB
Median (IQR)
concentration was
1.05 µg/L (0.98).
Cd concentrations were not significantly
associated with any outcomes.
Nishijo et al.,
200417
Toyama, Japan
N=55
Mean (range) concentration
was 9.29 (1.43-39.6)
nmol/L.
Cd concentrations were correlated with BL (r = -
0.337, p<0.01). In regression analyses, a 1
nmol/L increase in Cd concentration was
associated with a 0.59 cm decrease in BL (SE =
115
Samples collected at 30-32
weeks gestation
BL
0.277, p = 0.038), after adjusting for gestational
age and maternal weight gain at 30-32 weeks.
Zhang et al. 200418 Da-ye city, Hubei province,
China
n=44
Samples collected between
1 and 72 hours before
delivery.
PTB, BL
Concentrations ranged from
0.80 to 25.20 µg/L, with a
median concentration of
1.72 µg/L.
No associations between blood cadmium and
PTB or birth length were found.
Salpietro et al
200219
Messina, Italy
n=45
Samples collected during
prenatal care (weeks
gestation not reported) or at
delivery
BW
Mean (SD) concentration
was 119 ng/L (75).
Cd concentrations were significantly correlated
with BW (ρ= - 0.546; p=0.0003).
Odland et al.,
199920
Kola Peninsula of Russia
and Norway
n=262
Samples collected on day
of delivery
BW
Median (range)
concentrations were 2.2
(0.5-35.2) nmol/L among
women from Russia
(n=148) and 1.8 (0.5-26.9)
nmol/L among women
from Norway (n=114)
No significant associations were found between
Cd concentrations and BW.
116
Urine
Zhang et al. 201821 Guiyu (e-waste area) and
Haojing, China
n=237 (Guiyu), 212
(Hoajing)
Samples collected on day
of delivery
BW, BL, HC, apgar score 1
min and 5 min
Median (25th, 75th
percentiles) were reported
by location and sex:
Males: Haojiang and
Guiyu: 0.67 (0.31 1.05) and
0.92 (0.55, 1.66) µg/g
creatinine
Females: Haojiang and
Guiyu: 0.59 (0.29, 0.90)
and 1.00 (0.69, 1.77) µg/g
creatinine
Cd concentrations were significantly higher
among women in Guiyu and higher among
women giving birth to dgirls vs boys. For
women in Guiyu, living in a house near e-waste
site or highway was positively correlated with
Cd concentrations.
Among girls, an increase in 1 ug/g creatine Cd
led to effects of -9 g (95% CI: -75, -2 g) in BW,
-0.19 cm (95% CI: -0.36, -0.01 cm) in BL, and -
0.09 (95% CI: -0.15, -0.03) in apgar 1-min
score.
Guo et al. 201722 Jiangsu Province, China
n=1,073
Samples collected on day
of delivery
BW, BL, HC, PI
Median (25th, 75th
percentile) concentration
was 0.19 (0.08, 1.00) µg/g
creatinine
No significant effects were found between Cd
concentrations and outcomes.
Huang et al. 2017 23 Hubei province, China
n=408
Samples collected on day
of delivery
Preterm LBW (PLBW)
Mean or median
concentrations were not
reported. Authors provide
concentrations for l (<0.35
µg/g creatinine), medium
(0.35-0.70 µg/g creatinine),
high (>=0.70 µg/g
creatinine) exposure
groups.
Nested case-control study: controls were babies
≥37 weeks gestation, ≥2500g; cases were babies
<37 weeks, <2500g
An increased risk of PLBW was found for the
highest exposure group (OR = 2.25; 95% CI:
1.21, 4.17). When stratified by sex, significant
effects were only seen among girls OR= 5.90;
95% CI: 1.57, 22.23).
Wai et al. 201724 Ayeyarwady Division,
Myanmar
n=419
Median (IQR)
concentration was 0.86
(0.50-1.40) µg/g creatinine.
A 1-unit increase in Cd concentration was
associated with an increased risk for LBW (OR
= 1.10; 95% CI: 1.01, 1.21).
117
Samples collected in the
third trimester
LBW, PTB
Romano et al.
201625
Seattle, Washington
n=396
Samples collected at ~ 15
weeks gestation
BW, BL, HC, PI
Overall mean or median
concentrations were not
reported. Authors reported
concentrations by tertiles:
low (<0.29 μg/g creatinine),
medium (0.29–0.42 μg/g
creatinine), and high (≥0.43
μg/g creatinine)
No significant effects seen among all births.
When stratified by sex, each log unit increase in
Cd concentration was associated with a 0.47
(95% CI: 0.2, 0.74) cm decrease in BL. Among
boys, each log-unit decrease was associated with
a 0.63 (95% CI: 0.24, 1.01) kg/m3 decrease in
PI.
Yang et al. 201626 Wuhan, China
n=5,364
Samples collected on day
of delivery
BW, BL, GA, PTB, LBW,
SGA
GM (range) cd
concentration was: 0.55
(0.01-2.85) µg/g creatinine.
A log-unit increase in Cd concentration was
associated with a 0.77 day (95% CI: 0.39, 1.15
day) decrease in gestational age among all
births. Each log-unit increase in Cd
concentration was also associated with PTB (OR
= 1.78; 95% CI: 1.45, 2.19). No significant
effects seen for BW or BL. Interaction models of
sex were not significant.
Kippler et al.
201227
Rural Bangladesh
n=1,616
Samples collected at ~8
weeks gestation
BW, BL, HC, chest
circumference (CC)
Median concentration was
0.63 μg/L.
Cd concentrations were associated with BW (–
31.0 g; 95% CI: –59, –2.8 g) and HC (–0.15 cm;
95% CI: –0.27, –0.02 cm) per 1 µg/L increase.
When stratified by sex, significant effects were
only seen among girls: 1-μg/L increase in Cd
was associated with a 0.26-cm (95% CI: –0.43, –
0.088 cm) and 0.24-cm (95% CI: –0.44, –0.030
cm) decrease in girls’ HC and CC, respectively,
and a 45-g (95% CI: –82.5, 7.3 g) decrease in
BW. No evidence of dose-response effect was
seen in quantile regression analyses.
Shirai et al. 2010 28 Tokyo, Japan
n=78
Mean (SD) concentration
was 0.976 (0.891) µg/L.
A log-unit increase in Cd concentration was
associated with a -135 g (p=0.021) decrease in
BW.
118
Samples collected at
between 9 to 40 weeks
BW, BL, HC
Other media
Everson et al.
201629
Rhode Island, US
n=242
Toenail clippings collected
at a mean 2.8 months
gestation
SGA
Range was 0.00062 –
0.0846 µg/g.
A log-unit increase in Cd concentration was
associated with an increased risk of SGA (OR =
2.44; 95% CI: 1.53, 3.89).
aDefined as estimated fetal weight <3rd or <10th percentile with cerebro-placental ratio <5th and/or mean uterine artery pulsatility index >95th percentile
pathological Doppler.
119
References
1. Sabra S, Malmqvist E, Saborit A, Gratacós E, Gomez Roig MD. Heavy metals exposure
levels and their correlation with different clinical forms of fetal growth restriction. PLoS
ONE 2017; 12(10): e0185645.
2. Cheng LL. Critical Windows of Prenatal Exposure to Cadmium and Size at Birth.
International Journal of Environmental Research and Public Health 2016; 14(1): 58.
3. Luo Y, McCullough LE, Tzeng J-Y, et al. Maternal blood cadmium, lead and arsenic
levels, nutrient combinations, and offspring birthweight. BMC public health 2017; 17: 354.
4. Taylor CM, Golding J, Emond AM. Moderate Prenatal Cadmium Exposure and Adverse
Birth Outcomes: a Role for Sex-Specific Differences? Paediatric and Perinatal
Epidemiology 2016; 30(6): 603-11.
5. Wang H, Liu L, Hu Y-F, et al. Maternal serum cadmium level during pregnancy and its
association with small for gestational age infants: a population-based birth cohort study.
Scientific Reports 2016; 6: 22631.
6. Bloom MS, Buck Louis GM, Sundaram R, Maisog JM, Steuerwald AJ, Parsons PJ. Birth
outcomes and background exposures to select elements, the Longitudinal Investigation of
Fertility and the Environment (LIFE). Environmental research 2015; 138: 118-29.
7. Hu X, Zheng T, Cheng Y, et al. Distributions of Heavy Metals in Maternal and Cord Blood
and the Association with Infant Birth Weight in China. The Journal of Reproductive
Medicine 2015; 60(1-2): 21-9.
8. Röllin HB, Kootbodien T, Channa K, Odland JØ. Prenatal Exposure to Cadmium, Placental
Permeability and Birth Outcomes in Coastal Populations of South Africa. PLoS One 2015;
10(11): e0142455.
9. Thomas S, Arbuckle TE, Fisher M, Fraser WD, Ettinger A, King W. Metals exposure and
risk of small-for-gestational age birth in a Canadian birth cohort: The MIREC study.
Environmental Research 2015; 140(0): 430-9.
10. Vidal AC, Semenova V, Darrah T, et al. Maternal cadmium, iron and zinc levels, DNA
methylation and birth weight. BMC Pharmacology and Toxicology 2015; 16(1): 20.
11. Al-Saleh I, Shinwari N, Mashhour A, Rabah A. Birth outcome measures and maternal
exposure to heavy metals (lead, cadmium and mercury) in Saudi Arabian population. Int J
Hyg Environ Health 2014; 217(2-3): 205-18.
120
12. Johnston JE, Valentiner E, Maxson P, Miranda ML, Fry RC. Maternal cadmium levels
during pregnancy associated with lower birth weight in infants in a North Carolina cohort.
PLoS One 2014; 9(10): e109661.
13. Sun H, Chen W, Wang D, Jin Y, Chen X, Xu Y. The effects of prenatal exposure to low-
level cadmium, lead and selenium on birth outcomes. Chemosphere 2014; 108: 33-9.
14. Ikeh-Tawari EP, Anetor JI, Charles-Davies MA. Cadmium Level in Pregnancy, Influence
on Neonatal Birth Weight and Possible Amelioration by Some Essential Trace Elements.
Toxicology International 2013; 20(1): 108-12.
15. Menai M, Heude B, Slama R, et al. Association between maternal blood cadmium during
pregnancy and birth weight and the risk of fetal growth restriction: The EDEN mother–
child cohort study. Reproductive Toxicology 2012; 34(4): 622-7.
16. Lin C-MCM. Does prenatal cadmium exposure affect fetal and child growth? Occupational
and Environmental Medicine (London, England) 2011; 68(9): 641-6.
17. Nishijo M, Tawara K, Honda R, Nakagawa H, Tanebe K, Saito S. Relationship Between
Newborn Size and Mother's Blood Cadmium Levels, Toyama, Japan. Archives of
Environmental Health: An International Journal 2004; 59(1): 22-5.
18. Zhang Y-LYL. Effect of environmental exposure to cadmium on pregnancy outcome and
fetal growth: a study on healthy pregnant women in China. Journal of environmental
science and health Part A, Toxic/hazardous Substances & Environmental Engineering
2004; 39(9): 2507-15.
19. Salpietro CD, Gangemi S, Minciullo PL, et al. Cadmium concentration in maternal and
cord blood and infant birth weight: a study on healthy non-smoking women. Journal of
Perinatal Medicine 2002; 30(5): 395-9.
20. Odland J, Nieboer E, Romanova N, Thomassen Y, Lund E. Blood lead and cadmium and
birth weight among sub-arctic and arctic populations of Norway and Russia. Acta
Obstetricia et Gynecologica Scandinavica 1999; 78(10): 852-60.
21. Zhang Y, Xu X, Chen A, et al. Maternal urinary cadmium levels during pregnancy
associated with risk of sex-dependent birth outcomes from an e-waste pollution site in
China. Reproductive Toxicology 2018; 75(Supplement C): 49-55.
22. Guo J, Wu C, Qi X, et al. Adverse associations between maternal and neonatal cadmium
exposure and birth outcomes. Science of The Total Environment 2017; 575(Supplement
C): 581-7.
23. Huang K, Li H, Zhang B, et al. Prenatal cadmium exposure and preterm low birth weight
in China. Journal Of Exposure Science And Environmental Epidemiology 2017; 27: 491-
6.
121
24. Wai MK, Mar O, Kosaka S, Umemura M, Watanabe C. Prenatal Heavy Metal Exposure
and Adverse Birth Outcomes in Myanmar: A Birth-Cohort Study. International Journal of
Environmental Research and Public Health 2017; 14(11).
25. Romano ME, Enquobahrie DA, Simpson C, Checkoway H, Williams MA. Maternal body
burden of cadmium and offspring size at birth. Environmental Research 2016; 147: 461-8.
26. Yang J, Huo W, Zhang B, et al. Maternal urinary cadmium concentrations in relation to
preterm birth in the Healthy Baby Cohort Study in China. Environment International 2016;
94(Supplement C): 300-6.
27. Kippler M, Tofail F, Gardner R, et al. Maternal Cadmium Exposure during Pregnancy and
Size at Birth: A Prospective Cohort Study. Environmental Health Perspectives 2012;
120(2): 284-9.
28. Shirai S, Suzuki Y, Yoshinaga J, Mizumoto Y. Maternal exposure to low-level heavy
metals during pregnancy and birth size. Journal of Environmental Science and Health, Part
A 2010; 45(11): 1468-74.
29. Everson TM, Armstrong DA, Jackson BP, Green BB, Karagas MR, Marsit CJ. Maternal
cadmium, placental PCDHAC1, and fetal development. Reproductive Toxicology
(Elmsford, NY) 2016; 65: 263-71.
122
Appendix B. Supplemental material for chapter two
Figure B.1. Locations of control and intervention homes at time of at enrollment
*Home locations have been randomly "jittered" to protect the confidentiality of participants.
Text B.1 Description of Dylos data preparation
123
We conducted 911 one-week sampling events using Dylos particle counters, representing over
1.7 million 5-minute averaged concentrations. We completed several quality control and data
cleaning steps prior to data analysis. First, we removed measurements if more than 10% of the 5-
minute concentrations were 0 particles/cm3 for a given sampling event (n = 57). Next, we removed
measurements made by faulty Dylos monitors, as identified by co-location tests (n = 342). We
conducted co-location tests using all of our 42 Dylos monitors to identify faulty monitors. These
tests were performed prior to data collection, and again in July 2014, February 2015, and March
2016. We operated monitors operated side by side for a minimum of 24-hours during which 5-min
average particle counts were collected. The data were downloaded and the median concentration
across all Dylos units was calculated for each 5-minute period. For each Dylos unit, we then
regressed measured concentrations against the median concentrations. Since we expected
measurements made with functional Dylos monitors to agree with the median concentration across
all monitors (i.e., R2 ≈ 1; slope ≈ 1, and intercept ≈ 0), Dylos monitors with co-location regression
results that met any of the following criteria were considered faulty: 1) R2 < 0.80; 2) slope < 0.80
or > 1.20; or 3) intercept > 163,830 particles (equivalent of ~ 10 µg/m3 PM2.5). We identified 18
faulty monitors, and removed all field measurements made with those monitors from our dataset.
In total, we removed data from 342 sampling events. Finally, we excluded 52 measurements where
the Dylos monitors recorded a concentration for less than 50% of the one-week sampling event
and five measurements with incorrect filenames that could not be matched to an UGAAR
participant. We concluded that relative humidity (RH) did not influence Dylos particle counts
based on (i) the low RH values measured in homes, and (ii) the weak relationship between RH and
Dylos particle counts. In total, 175,467 5-min RH measurements were collected in 85 homes, with
a geometric mean of 27.9 % (95 % CI: 27.9, 28.0 %). Regression analysis of 16,481 matched 5-
minute RH and Dylos particle count measurements showed a weak relationship (R2 = 0.13).
Removal of data collected from participants who were lost to follow up left a total of 447 one-
week Dylos measurements for analysis. We assessed baseline housing, personal, and behavioral
characteristics among participants with no measurements and participants with one or two
measurements using Fisher’s exact tests and Mann-Whitney tests as appropriate. Participants with
no measurements spent less time at home in early pregnancy (15.7 hours/day, 95% CI: 15.1, 16.2
hrs/day) compared with participants with one or two measurements (16.3 hours/day, 95% CI: 15.9,
16.8 hrs/day, p = 0.02). We found no other significant differences between these groups (Table
124
S2.1). We concluded participants and homes included in our analysis are representative of the full
UGAAR cohort.
125
Table B.1. Summary of household, personal, behavioral, and intervention-related
characteristics for participants with zero and one or two one-week Dylos
measurements
Participants with zero
one-week Dylos
measurements
(n = 170)
Participants with one or
two one-week Dylos
measurements
(n = 342)
p-value
GM (95 % CI) or
N %
GM (95 % CI) or
N %
Housing characteristics
Total home area (m2) 52.0 (48.1, 56.2) 91 54.0 (50.9, 57.3) 98 0.76
Not recorded 9 2
Age of home (years) 10.6 (8.4, 13.4) 66 10.9 (9.3, 12.9) 71 0.74
Not recorded 34 29
Window opening in winter
Open < half the month 82 48 168 49 0.85
Open ≥ half the month 86 51 169 50
Not recorded 2 1 5 1
Window opening in summer
Open < half the month 21 12 38 11 0.66
Open ≥ half the month 147 86 301 88
Not recorded 2 1 3 1
Used a non-UGAAR air cleaner
No 160 94 319 93 0.63
Yes 5 3 15 5
Not recorded 5 3 8 1
Personal and behavioral characteristics
Time spent indoors at
home in early pregnancy
(hours/day) 15.7 (15.1, 16.2) 74 16.3 (15.9, 16.8) 78
0.02
Not recorded 26 22
Time spent indoors at
home in late pregnancy
(hours/day) 15.9 (15.3, 16.6) 50 15.6 (14.9, 16.3) 55
0.82
Not recorded 50 45
Smoked at any point pregnancy
No 157 92 309 90 0.30
Yes 10 6 30 9
Not recorded 3 2 3 1
Lived with a smoker in at any time in pregnancy
No 81 48 168 49 0.98
Yes 81 48 169 50
Not recorded 8 4 5 1
126
Intervention-related characteristics
Treatment group
Control 89 52 164 48 0.40
Intervention 81 48 178 52
Week of pregnancy at
enrollment 9.8 (9.3, 10.2) 100 10.0 (9.7, 10.3) 100
0.35
Season of enrollment
Winter (Dec-Feb) 63 37 104 30 0.43
Spring (Mar-May) 46 27 95 28
Summer (Jun-Aug) 20 12 42 12
Fall (Sep-Nov) 41 24 101 30
127
Text B.2. Description of Dylos particle count conversion to PM2.5 concentrations
Particle counts were converted to PM2.5 concentrations using a conversion equation based on co-
located Dylos and gravimetric PM2.5 measurements. In total, 100 gravimetric filter samples were
collected, of which 10 were field blanks, and two were duplicate samples collected side-by-side in
the same home. The mean (standard deviation) of blanks filters was 0.01 (0.02) µg indicating
minimal contamination of filters. The duplicate samples showed excellent agreement with respect
to PM2.5 concentrations, with values of 15.4 and 16.0 µg/m3. Forty-one gravimetric samples were
discarded due to incomplete data or differences of > 10 % in start and stop pump flow rates. The
remaining 49 samples were matched with Dylos particle count data, resulting in 23 valid pairs of
Dylos measurements and gravimetric PM2.5 samples. To establish the conversion equation, Dylos
particle counts were regressed on gravimetric PM2.5 measurements, giving the following equation:
PM2.5 (µg/m3) = 1.88 + 1.34*(Particle Count, particles/cm3), R2 = 0.94 (Figure S3.1). This
equation was used to convert all one-week particle counts to PM2.5 concentrations.
128
Table B.2. Summary of household, personal, behavioral, and intervention-related
characteristics for participants who were lost to follow up and those who
remained in the study
Lost to follow up
(n = 28)
Remained in study
(n = 512)
p-value
GM (95 % CI) or
N %
GM (95 % CI) or
N %
Housing characteristics
Total home area (m2) 56.2 (44.8, 70.5) 64 53.4 (50.9, 55.9) 95 0.79
Not recorded 36 5
Age of home (years) 14.2 (8.5, 23.7) 50 10.9 (9.5, 12.5) 69 0.71
Not recorded 50 31
Window opening in winter
Open < half the month 10 36 248 48 0.98
Open ≥ half the month 11 39 255 50
Not recorded 7 25 9 2
Window opening in summer
Open < half the month 3 11 59 11 0.73
Open ≥ half the month 18 64 448 88
Not recorded 7 25 5 1
Used a non-UGAAR air
cleaner
0.05
No 19 68 479 94
Yes 2 7 20 4
Not recorded 7 25 13 2
Outdoor PM2.5 (µg/m3) at
first home measurement 54.3 (50.5, 58.4) 68 55.0 (51.6, 58.6) 97
0.80
Not available 32 3
Personal and behavioral characteristics
Time spent indoors at home
in early pregnancy
(hours/day) 15.1 (13.4, 17.0) 54 16.1 (15.8, 16.4) 76
0.20
Not recorded 46 24
Lived with a smoker in early pregnancy
No 20 71 464 90 0.60
Yes 1 4 40 8
Not recorded 7 25 8 2
Smoking occurred in the home at early in pregnancy
No 10 36 264 51 0.85
Yes 11 39 230 45
Not recorded 7 25 18 4
Intervention-related characteristics
Treatment group
Control 19 68 253 49 0.08
129
Intervention 9 32 259 51
Week of pregnancy at
enrollment 9.2 (8.2, 10.3)
10
0 9.9 (9.7, 10.2)
10
0
0.15
Season of enrollment
Winter (Dec-Feb) 9 32 167 33 0.12
Spring (Mar-May) 13 46 142 28
Summer (Jun-Aug) 1 4 62 12
Fall (Sep-Nov) 5 18 141 28
130
Figure B.2. Relationship between one-week outdoor and indoor PM2.5 concentrations
for control and intervention homes
131
Table B.3. Summary of previous studies investigating air cleaner effectiveness in residences over periods of 6 to 12 months
Study PM source of
interest Study design Reduction in residential indoor particulate matter
McNamara
et al. 2017 1
Wood smoke Homes were randomized to one of two groups
for 12 months: (i) the high efficiency filtration
(mechanical filter with > 90% efficiency for
capturing particles 1-3 µm; n = 25) group; or (ii)
the low efficiency filtration (fiberglass filter; n =
23) group.
66 %
Median (range) PM2.5 concentrations were 22.0 µg/m3
in low and 5.7 µg/m3 in high efficiency filtration
homes at the 12-month follow up period,
respectively.
Batterman
et al. 2012 2
No specific
source
Homes were randomized to one of three groups
for 3-4 consecutive seasons: (i) control (n=37);
(ii) HEPA cleaner (n=47); or (iii) HEPA cleaner
and air conditioner (n=42)
45 %
Mean (sd) PM2.5 concentrations were 21.4 (18.1)
µg/m3 at baseline and 11.8 (10.9) µg/m3 over period
when air cleaners were used over 3-4 consecutive
seasons, in homes receiving HEPA cleaners.
Butz et al.
2011 3
Second hand
smoke
Homes were randomized to one of three groups
for 6 months: (i) control (n=44); (ii) 2 HEPA
cleaners (n=41); or (iii) 2 HEPA cleaners plus
home visits from a health coach (n=41)
47 %
Mean (sd) PM2.5 concentrations were 33.9
(26.4) µg/m3 at baseline and 17.9 (15.2) µg/m3 at the
6-month follow up period, in homes receiving HEPA
cleaners.
Lanphear et
al. 2011 4
Second hand
smoke
Homes were randomized to one of two groups
for 12 months: (i) 2 sham air cleaners (n=115);
or (ii) 2 HEPA filter air cleaners (n=110)
25-38%
Mean particle counts (> 0.3 um) were 4.0 x 106/ft3,
2.5 x 106/ft3, and 3.0 x 106/ft3 at baseline, 6-month
and 12-month follow up periods.
References
1. McNamara ML, Thornburg J, Semmens EO, Ward TJ, Noonan CW. Reducing indoor air
pollutants with air filtration units in wood stove homes. Science of The Total Environment
2017; 592: 488-94.
2. Batterman S, Du L, Mentz G, et al. Particulate matter concentrations in residences: an
intervention study evaluating stand-alone filters and air conditioners. Indoor Air 2012; 22(3):
235-52.
3. Butz AM, Matsui EC, Breysse P, et al. A randomized trial of air cleaners and a health coach
to improve indoor air quality for inner-city children with asthma and secondhand smoke
exposure. Archives of Pediatrics & Adolescent Medicine 2011; 165(8): 741-8.
4. Lanphear BP, Hornung RW, Khoury J, Yolton K, Lierl M, Kalkbrenner A. Effects of HEPA
air cleaners on unscheduled asthma visits and asthma symptoms for children exposed to
secondhand tobacco smoke. Pediatrics 2011; 127(1): 93-101.
133
Appendix C. Supplemental material for chapter three
Figure C.1. Summary of all data collected in the Ulaanbaatar Gestation and Air
Pollution Research (UGAAR) study
134
Table C.1. Summary of baseline characteristics among those lost to follow up and those
that remained in the study
Lost to follow up
(n=28)
Remained in study
(n=512) p-
valuea
Median
(25th, 75th percentile)
or N (%)
Median
(25th, 75th percentile)
or N (%)
Mother’s age at enrollment,
weeks 28.5 (25.0, 34.0) 29.5 (25.0, 33.0)
0.74
Gestational age at enrollment,
weeks 10.0 (8.0,12.0) 10.0 (8.0,13.0)
0.17
Household income
< 800,000 Tugriksb 6 (21) 170 (34) 0.20
≥ 800, 000 Tugriks 12 (43) 335 (65)
Not reported, N (%) 10 (36) 7 (1)
Mother’s education
Completed university 17 (61) 411 (80) 0.50
Did not complete university 4 (14) 61 (12)
Not reported, N (%) 7 (25) 40 (8)
Marital status
Married/common-law 16 (57) 436 (85) 0.52
Not married/common-law 4 (14) 75 (15)
Not reported, N (%) 8 (29) 1 (0)
Worked/volunteered outside the home
No 6 (21) 165 (33) 0.81
Yes 15 (54) 340 (66)
Not reported, N (%) 7 (25) 7 (1)
Parity
0 2 (7) 46 (9) 0.78
1 6 (21) 192 (38)
≥2 5 (18) 118 (23)
Not reported, N (%) 15 (54) 156 (30)
Time since last pregnancy,
months 25 (11, 55) 31 (15, 60)
0.13
Not reported, N (%) 19 (68) 231 (45)
Previous poor pregnancy outcomec
No 6 (21) 104 (20) 0.16
Yes 2 (7) 114 (22)
Not reported, N (%) 20 (72) 294 (58)
Pre-pregnancy BMI, kg/m2 21.2 (19.7, 24.0) 21.5 (19.8, 24.0) 0.48
Not reported, N (%) 7 (25) 35 (7)
135
Time spent at home in early
pregnancy, hrs/day 16.0 (14.0, 18.7) 16.1 (14.4, 18.8)
0.20
Not reported, N (%) 13 (46) 121 (24)
Smoked at any time in pregnancy, N (%)
No 20 (71) 464 (91) 0.95
Yes 1 (4) 40 (8)
Not reported 7 (25) 8 (1)
Lived with a smoker at any time in pregnancy, N (%)
No 10 (36) 266 (52) 0.90
Yes 11 (39) 230 (45)
Not reported 7 (25) 16 (3)
Alcohol consumption at any time in pregnancy, N (%)
No 13 (46) 320 (63) 0.90
Yes 7 (25) 165 (32)
Not reported 8 (29) 27 (5)
Drug use at any time in pregnancy, N (%)
No 21 (75) 493 (96) 0.96
Yes 0 (0) 3 (1)
Not reported 7 (25) 16 (3) ap-values were generated using non-parametric Wilcoxon rank tests for continuous outcomes and Fisher’s exact tests
for categorical outcomes. bAt the time of data collection 800,000 Tugriks was the equivalent of approximately $360 US. cPrevious poor outcome includes spontaneous abortion, still birth, low birthweight, macrosomia, ectopic pregnancy,
birth defect, and intrauterine growth restriction.
136
Table C.2. Summary of sensitivity analyses conducted to investigate intervention effects
on birth weight
aBirths occurring ≥ 37 weeks gestation. bDefined as observations that exceeded the WHO fetal growth chart 95th percentiles of birth weight for gestational age
and sex by ≥20%
Sensitivity Analysis All births Term birthsa
Participants analyzed according to the
intervention they received (i.e. “per
protocol”)
-15 (-117, 87)
n=463 69 (-13, 150)
n=429
Excluding two observations identified as
having potential errors in gestational age
or birth weight (both observations were
preterm)
15 (-88, 117)
n=461
------
Excluding five neonatal deaths 37 (-63, 137)
n=459
86 (4, 168)
n=428
Excluding participants who reported
smoking at any time during pregnancy 5 (-101, 112)
n=424
76 (-10, 162)
N=393
Model adjusted for anemia status 15 (-87, 118)
n=463
85 (3, 167)
n=429
Model adjusted for preterm birth 84 (-1, 170)
n=463 ------
137
Appendix D. Supplemental material for chapter four
Table D.1. Summary of characteristics for UGAAR participants who had a live birth
and (i) did not have a blood cadmium measure collected and (ii) had a blood
cadmium measure collected.
Participants with no
blood cadmium
measurement
(n = 86)
Participants with a
blood cadmium
measurement
(n=374)
p-value
Median (25th, 75th
percentile)
Or N (%)
Median (25th, 75th
percentile)
Or N (%)
MATERNAL
Age at enrollment (yrs) 29 (25, 34) 29 (25, 33) 0.76
Gestational age at enrollment
(weeks)
12 (10, 13) 10 (8, 12) 0.003
Season of enrollment
Winter (December, January,
February)
62 (72) 86 (23) 0.0001
Spring (March, April, May) 14 (16) 114 (30)
Summer (June, July, August) 2 (2) 54 (14)
Fall (September, October,
November)
8 (9) 120 (32)
Monthly household income
< 600,000 Tugriksa 18 (21) 92 (25) 0.8
600,000 to <1,200,000 Tugriks 25 (29) 105 (28)
≥ 1,200,000 Tugriks 39 (45) 163 (44)
Missing 4 (5) 14 (4)
Maternal education
Completed university 67 (78) 302 (81) 0.22
Did not complete university 14 (16) 42 (11)
Missing 5 (6) 30 (8)
Marital status
Married/common-law 68 (79) 320 (86) 0.14
Not married/common-law 18 (21) 54 (14)
Missing 0 (0) 0 (0)
Parity
0 6 (7) 39 (10) 0.59
1 32 (37) 141 (38)
≥2 21 (24) 82 (22)
Missing 27 (31) 112 (30)
Pre-preg BMI (GM, kg/m2) 21.9 (20, 23.9) 21.5 (19.5, 24) 0.36
Missing 3 (3) 26 (7)
Tobacco smoke exposures
Smoked at any time in
pregnancy
No 76 (88) 344 (92) 0.31
138
Yes 9 (10) 27 (7)
Missing 1 (1) 3 (1)
Lived with a smoker at any time
in pregnancy
No 41 (48) 185 (49) 0.79
Yes 43 (50) 182 (49)
Missing 2 (2) 7 (2)
Health during current
pregnancy
Anemia
No 69 (80) 304 (81) 0.83
Yes 17 (20) 70 (19)
Diabetes
No 86 (100) 374 (100) NA
Yes 0 (0) 0 (0)
Gestational diabetes
No 86 (100) 374 (100) NA
Yes 0 (0) 0 (0)
Hypertension
No 84 (98) 352 (94) 0.18
Yes 2 (2) 22 (6)
Gestational hypertension
No 63 (73) 327 (87) 0.01
Yes 0 (0) 32 (9)
Missing
TORCH infections
No 71 (83) 367 (98) 0.78
Yes 1 (1) 7 (2)
Missing 14 (16) 0 (0)
NEWBORN
Birth weight (g) 3450 (3000, 3900) 3500 (3200, 3800) 0.44
Gestational age at birth (weeks) 39.3 (38.5, 40) 39.5 (38.5, 40) 0.01
Birth length (cm) 50.5 (50, 52) 51 (50, 52) 0.15
Head circumference (cm) 35 (34, 36) 35 (34, 36) 0.13
Ponderal index (g/cm3) 2.7 (2.5, 2.8) 2.6 (2.5, 2.8) 0.59
Low birth weight 8 (9) 15 (4) 0.04
Small for gestational age 6 (7) 27 (7) 0.94
Preterm birth 15 (17) 19 (5) 0.001 aAt the time of data collection, 600,000 Tugriks was the equivalent of approximately $243 USD
139
Table D.2. Estimated effects of a doubling of maternal blood cadmium concentrations on fetal growth outcomes, stratified
by sex
Girls Boys
Crude Adjusteda Crude Adjusteda
Outcome
Type of
effect
estimateb
All births
n = (169)
Term births
n = (160)
All births
n = (143)
Term births
n = (136)
All births
n = (205)
Term
births
n = (195)
All births
n = (181)
Term
births
n = (175)
Birth weight,
g
Mean
difference
(95% CI)
-115
(-210, -19)
-116
(-200, -32)
-103
(-198, -8)
-116
(-205, -27)
-39
(-126, 47)
-62
(-139, 15)
-63
(-144, 18)
-58
(-138, 22)
Birth length,
cm
-0.33
(-0.82, 0.16)
-0.17
(-0.55, 0.21)
-0.15
(-0.59, 0.28)
-0.18
(-0.60, 0.25)
-0.13
(-0.52, 0.26)
-0.23
(-0.57, 0.12)
-0.15
(-0.52, 0.22)
-0.17
(-0.54,
0.21)
Head
circumference,
cm
-0.24
(-0.57, 0.08)
-0.22
(-0.51, 0.07)
-0.27
(-0.59, 0.05)
-0.31
(-0.61, 0.00)
0.12
(-0.18, 0.42)
0.02
(-0.24, 0.27)
0.06
(-0.20, 0.31)
0.07
(-0.2, 0.33)
Ponderal
index, g/cm3
-0.03
(-0.09, 0.04)
-0.06
(-0.11, -0.02)
-0.06
(-0.13, 0.004)
-0.07
(-0.12, -0.02)
-0.01
(-0.05, 0.04)
-0.01
(-0.05, 0.04)
-0.03
(-0.08, 0.019)
-0.02
(-0.07,
0.03)
Low
birthweight Odds
ratio
(95% CI)
1.38
(0.59, 3.25)
2.06
(0.64, 6.61)
1.13
(0.39, 3.33)
1.16
(0.39, 3.49)
0.54
(0.2, 1.45)
1.95
(0.49, 7.79)
1.21
(0.46, 3.14)
1.00
(0.38, 2.64)
Small for
gestational age
1.83
(0.95, 3.50)
2.30
(1.13, 4.66)
1.85
(0.90, 3.81)
2.03
(0.95, 4.33)
1.14
(0.58, 2.23)
0.82
(0.34, 1.96)
1.21
(0.6, 2.44)
0.84
(0.35, 2.00)
Preterm birth 0.63
(0.26, 1.50)
1.75
(0.70, 4.38)
0.70
(0.31, 1.58)
1.61
(0.68, 3.8)
aModels of birth weight, birth length, head circumference, ponderal index and low birth weight were adjusted for maternal age, monthly household income, pre-pregnancy BMI,
anemia status, gestational age and gestational age squared, living with a smoker in late pregnancy, ger density, and intervention status. Models of small for gestational age were
adjusted for all same list of variables, except for gestational age, gestational age squared, and sex, and models of preterm birth were adjusted for the same list of variables, except for
gestational age and gestational age squared. bPer doubling of blood cadmium concentrations.
140
Table D.3. Estimated effects of a doubling of maternal blood cadmium concentrations on fetal growth outcomes, stratified
by living with a smoker in late pregnancy.
Did not live with a smoker in late pregnancy Lived with a smoker in late pregnancy
Crude Adjusteda Crude Adjusteda
Outcome
Type of
effect
estimateb
All births
n = (193)
Term births
n = (185)
All births
n = (176)
Term births
n = (169)
All births
n = (160)
Term births
n = (154)
All births
n = (148)
Term births
n = (142)
Birth weight,
g
Mean
difference
(95% CI)
-56
(-166, 54)
-76
(-181, 28)
-44
(-148, 59)
-33
(-134, 67)
-104
(-187, -22)
-89
(-160, -18)
-108
(-180, -35)
-108
(-179, -38)
Birth length,
cm
-0.30
(-0.86, 0.25)
-0.18
(-0.65, 0.29)
-0.16
(-0.66, 0.34)
-0.05
(-0.53, 0.42)
-0.18
(-0.55, 0.19)
-0.15
(-0.46, 0.16)
-0.13
(-0.46, 0.2)
-0.15
(-0.48, 0.18)
Head
circumference,
cm
0.01
(-0.36, 0.37)
-0.01
(-0.34, 0.33)
0.04
(-0.28, 0.37)
0.07
(-0.25, 0.39)
-0.11
(-0.37, 0.15)
-0.08
(-0.31, 0.15)
-0.16
(-0.4, 0.08)
-0.16
(-0.40, 0.08)
Ponderal
index, g/cm3
0.03
(-0.04, 0.09)
-0.03
(-0.08, 0.02)
0.002
(-0.06, 0.069)
-0.01
(-0.07, 0.04)
-0.05
(-0.1, -0.01)
-0.05
(-0.09, -0.002)
-0.069
(-0.12, -0.023)
-0.06
(-0.11, -0.02)
Low
birthweight Odds
ratio
(95% CI)
0.43
(0.14, 1.35)
0.99
(0.15, 6.39)
1.13
(0.33, 3.91)
0.77
(0.20, 2.98)
1.78
(0.82, 3.87)
2.40
(0.79, 7.29)
1.32
(0.58, 3.05)
1.00
(0.43, 2.35)
Small for
gestational age
1.09
(0.48, 2.45)
1.16
(0.49, 2.74)
0.92
(0.39, 2.18)
0.98
(0.40, 2.39)
2.19
(1.15, 4.17)
2.16
(1.04, 4.48)
1.91
(1.02, 3.55)
1.67
(0.84, 3.31)
Preterm birth 0.65
(0.23, 1.87)
1.47
(0.52, 4.11)
1.33
(0.59, 2.99)
1.41
(0.63, 3.15)
aModels of birth weight, birth length, head circumference, ponderal index and low birth weight were adjusted for maternal age, monthly household income, pre-pregnancy BMI,
anemia status, sex of the baby, gestational age and gestational age squared, ger density, and intervention status. Models of small for gestational age were adjusted for all same list of
variables, except for gestational age and sex, and models of preterm birth were adjusted for the same list of variables, except for gestational age. bPer doubling of blood cadmium concentrations.
141
Table D.4. Estimated effects of a doubling of maternal blood cadmium concentrations on fetal growth outcomes, stratified
by ger density surrounding participants’ apartments
Birth weight,
g
Birth length,
cm
Head
circumferenc
e, cm
Ponderal
index, g/cm3
Low
birthweightc
Small for
gestational
age
Preterm
birth
Mean difference (95% CI) Odds ratio (95% CI)
< 3.5 gers
per
hectare
Crude All
births
n = 127 -43 (-164, 78)
0.05 (-0.48, 0.58)
0.02 (-0.39, 0.43)
-0.04 (-0.09, 0.02)
0.56 (0.16, 1.98)
1.18 (0.5, 2.75)
0.48 (0.17, 1.35)
Term
births
n = 118 -82
(-179, 14)
-0.06
(-0.48, 0.35)
-0.07
(-0.43, 0.28)
-0.05
(-0.1, 0.003)
1.49
(0.58, 3.82)
Adjusteda All
births
n = 108 -35 (-143, 73)
0.19 (-0.32, 0.69)
0.11 (-0.29, 0.51)
-0.05 (-0.11, 0.01)
1.02 (0.31, 3.32)
1.18 (0.89, 1.58)
1.75 (0.6, 5.13)
Term
births
n = 103 -52
(-154, 51)
0.13
(-0.37, 0.63)
0.07
(-0.32, 0.47)
-0.06
(-0.12, 0.01)
1.40
(0.99, 1.98)
3.5 - 4.5
gers per
hectare
Crude All
births
n = 130 -97 (-213, 18)
-0.24 (-0.75, 0.26)
-0.03 (-0.4, 0.33)
-0.04 (-0.1, 0.02)
2.69 (0.88, 8.25)
1.9 (0.92, 3.91)
2.37 (0.43, 13.13)
Term
births
n = 129 -87
(-199, 26)
-0.2
(-0.69, 0.3)
-0.001
(-0.36, 0.36)
-0.04
(-0.09, 0.019)
2.76
(0.8, 9.49)
1.85
(0.88, 3.87)
Adjusteda All
births
n = 118 -131
(-235, -28)
-0.32
(-0.76, 0.13)
-0.15
(-0.47, 0.17)
-0.07
(-0.13, -
0.0001)
1.88
(0.66, 5.39)
1.85
(0.88, 3.89)
1.18
(0.42, 3.32)
Term
births
n = 117 -125 (-230, -21)
-0.30 (-0.75, 0.15)
-0.12 (-0.45, 0.2)
-0.06 (-0.13, 0.01)
1.49 (0.57, 3.91)
1.86 (0.86, 4.02)
> 4.5 gers
per
hectare
Crude All
births
n = 115 -115
(-222, -8)
-0.56
(-1.18, 0.06)
-0.18
(-0.58, 0.23)
0.01
(-0.07, 0.09)
1.34
(0.51, 3.49)
1.22
(0.49, 3.05)
1.33
(0.6, 2.97)
Term
births
n = 107 -77 (-167, 14)
-0.33 (-0.77, 0.11)
-0.15 (-0.44, 0.13)
-0.003 (-0.05, 0.049)
0.32 (0.06, 1.62)
0.8 (0.25, 2.55)
Adjusteda All
births
n = 98 -102
(-206, 3)
-0.33
(-0.86, 0.21)
-0.22
(-0.55, 0.11)
-0.02
(-0.1, 0.06)
1.08
(0.44, 2.65)
1.22
(0.5, 3)
1.5
(0.63, 3.53)
Term
births
n = 91 -79
(-180, 22)
-0.21
(-0.69, 0.27)
-0.22
(-0.54, 0.1)
-0.02
(-0.08, 0.03)
1.00
(0.44, 2.26)
0.57
(0.16, 2.05)
aModels of birth weight, birth length, head circumference, ponderal index and low birth weight were adjusted for maternal age, monthly household income, pre-pregnancy BMI,
anemia status, sex of the baby, gestational age and gestational age squared, living with a smoker in late pregnancy, and intervention status. Models of small for gestational age were
adjusted for all same list of variables, except for gestational age and sex, and models of preterm birth were adjusted for the same list of variables, except for gestational age. bPer doubling of blood cadmium concentrations. cThere were no cases of low birth weight among term births in the intervention group
142
Table D.5. Estimated effects of a doubling of maternal blood cadmium concentrations on fetal growth outcomes, stratified
by intervention status
Control group Intervention group
Crude Adjusteda Crude Adjusteda
Outcome
Type of
effect
estimateb
All births
n = (173)
Term births
n = (169)
All births
n = (144)
Term births
n = (142)
All births
n = (201)
Term births
n = (186)
All births
n = (180)
Term births
n = (169)
Birth weight, g
Mean difference
(95% CI)
-51
(-142, 39)
-70
(-156, 16)
-95
(-174, -17)
-98
(-190, -6)
-88
(-182, 6)
-89
(-167, -11)
-100
(-184, -15)
-95
(-174, -17)
Birth length, cm -0.08
(-0.47, 0.31)
-0.16
(-0.54, 0.21)
-0.05
(-0.46, 0.36)
-0.08
(-0.49, 0.33)
-0.34
(-0.82, 0.14)
-0.18
(-0.54, 0.18)
-0.29
(-0.7, 0.12)
-0.23
(-0.62, 0.16)
Head
circumference,
cm
0.05
(-0.24, 0.33)
-0.01
(-0.29, 0.26)
-0.06
(-0.36, 0.25)
-0.06
(-0.37, 0.25)
-0.13
(-0.47, 0.21)
-0.14
(-0.42, 0.13)
-0.12
(-0.39, 0.15)
-0.13
(-0.39, 0.14)
Ponderal index, g/cm3
-0.02 (-0.07, 0.02)
-0.03 (-0.07, 0.02)
-0.06 (-0.11, -0.01)
-0.06 (-0.12, -0.01)
-0.004 (-0.06, 0.05)
-0.04 (-0.08, 0.01)
-0.03 (-0.09, 0.03)
-0.04 (-0.09, 0.01)
Low
birthweightc
Odds ratio (95% CI)
1.03
(0.41, 2.64)
1.55
(0.60, 4.04)
1.34
(0.48, 3.77)
1.57
(0.53, 4.69)
0.87
(0.37, 2.04)
----- 1.19
(0.40, 3.49)
-----
Small for gestational age
1.22 (0.67, 2.21)
1.21 (0.66, 2.22)
1.18 (0.89, 1.58)
1.40 (0.99, 1.98)
1.70 (0.81, 3.57)
1.85 (0.71, 4.84)
1.49 (0.72, 3.11)
1.51 (0.62, 3.65)
Preterm birth 0.60
(0.16, 2.23)
1.21
(0.40, 3.67)
0.77
(0.39, 1.54)
1.72
(0.82, 3.64)
aModels of birth weight, birth length, head circumference, ponderal index and low birth weight were adjusted for maternal age, monthly household income, pre-pregnancy BMI,
anemia status, sex of the baby, gestational age and gestational age squared, living with a smoker in late pregnancy, and ger density. Models of small for gestational age were adjusted
for all same list of variables, except for gestational age and sex, and models of preterm birth were adjusted for the same list of variables, except for gestational age. bPer doubling of blood cadmium concentrations. cThere were no cases of low birth weight among term births in the intervention group.